co») United States cz) Patent Application Publication (1 Pub. No.: US 2020/0054290 AI US 202000542901 JANG et al. (43) Pub, Date: Feb. 20, 2020 os) AGIB 0402 (2006.01) A61B 0488 (2006.01) GIB SI (2006.01) oy MSUNG ELECTRONICS €O., A018 5029 (2006.01) (72) Inventors: (73) Assign: (21) Appl. No, (2) File: Suwonesi (KR) Dac Geun JANG, Yoogin-si (KR): Ub Kun KWON, Hwascong-si (KR) Chang Soon PARK, Chungju-si (KR) SAMSUNG ELECTRONICS CO, LID, Suwousi (KR) 161814446 May 16, 2019 G0) Foreign Application Priority Data Aug. 20,2018 (KR) 10-2018-0096674 Publication Clasifcation (51) Inch Abi 5/00 2006.01) 461 S021 (200501) Giant (cai TECHN RFOCEST TOR ESTIMATING, BLOOD Het AGL 5/7278 (2013.01); ABIB S216 (2013.01), A6LB $/02125 (2013.01), 46112 02416 (2013.01): GIB S/0488 (2013.01): ‘AGLB 5/1102 (2013.01); AGLB $029 (2013.01); A6ZB $/0402 (2013.01) on ABSTRACT An apparatus for non-invasively estimating blood pressure is proved. According fo one embodiment, the spparats for estimating blood pressure may include a bio-signal sensor configured to. measure a bio-signal from a wser and a processor configured to extract a feature fom the bio-signal tan extraction time, acquire an offset based on 2 lative change value ofthe feature extracted atthe extraction time, relative ta reference valle ofthe Feature obtained 9 time of Calibration, and estimate blood pressure based on the rel five change valve ofthe extracted feature andthe acquited ofl AA: FXTRACT CO Ton EXTRACT TPR on ATEHIEST CHANG ‘ON ATLEAST ONE Of VARUES TSTIMATE ROOD PRESUA BRSEDON FIRST CHANGE VALUE, SECOND CHANGE VALDENANDONSEr TARCULRTESSCOND CHANGE VALUE RELATIVE T0.11ME SOE CALIGRATIONPatent Application Publication Feb. 20, 2020 Sheet 1 of 13 US 2020/0054290 AI FIG. 1 er |Patent Application Publication Feb. 20, 2020 Sheet 2 of 13. US 2020/0054290 AI FIG. 2 110 20 | 230 BIO-SIGNAL MEASURER PROCESSOR 230 ourpur INTERFACE eae: Posts tt tts sana eePatent Application Publication Feb. 20, 2020 Sheet 3 of 13. US 2020/0054290 AI FIG. 3A mmHg mmHy 160 ~120 100 BEFORE HIGH | ae G AFTER HIGH INTENSITY EXERCISE NEE EEC (STABLE CONDITION) eeePatent Application Publication Feb. 20, 2020 Sheet 4 of 13 ABP-(AfI+ AR) FIG. 3B ° SAMPLE OFESET ESTIMATION MODEL AMXAIZ US 2020/0054290 A1Patent Application Publication Feb. 20, 2020 Sheet 5 of 13. US 2020/0054290 AI FIG, 3C SAMPL _ OFFSET. go 5 && < oO z 3 ° 4 ° 9 ° ° 2 ° 3 ALVAHRPatent Application Publication Feb. 20, 2020 Sheet 6 of 13. US 2020/0054290 AI FIG. 4A FEATURE CHANGE ESTIMATOR OFFS ACQUIRER,Patent Application Publication Feb. 20, 2020 Sheet 7 of 13 US 2020/0054290 AI FIG. 4B ————>| CALIBRATOR 411 : i i \ \ ' i 1 i i = FEATURE 1] FEATURE | na EEATUI CHANGE | | EXTRACTOR ESTIMATOR ' \ \ \ i i 1 \ L QEFSET. ACQUIRERPatent Application Publication Feb. 20, 2020 Sheet 8 of 13. US 2020/0054290 AI > TIME re FIG. 5. Ty; Th Tsy: 0 AMPLITUDEPatent Application Publication Feb. 20, 2020 Sheet 9 of 13. US 2020/0054290 AI FIG. 5B AMPLITUDE(AU) 0 01 02 03 04 05 06 07 08 09 10 TIME (SECOND)Patent Application Publication Feb. 20, 2020 Sheet 10 of 13 US 2020/0054290 AI FIG. 6 START 610 EIVE REQUEST FOR ESTIMATING BLOOD PRESSURE, EXTRACT CO on CALCULATE FIRST CHANGE VALU! RELATIVE TO TIME OF CALIBRATION 642 CALCULATE SECOND CHANGE RELATIVE TC OF CALIBRATION CALCULATE THIRD eee ON ATLEAST ONE OF CHANGE VALUES. ESTIMATE BLOOD. PRESSURE BASED ON FIRST CHANGE VALUPatent Application Publication Feb. 20, 2020 Sheet 11 of 13. US 2020/0054290 AI FIG. 7 START RECHIVE REOUF FOR ESTIMATING BLOOD PRESSURE ACQUIRE BIO-SIGNAL EXTRACT PP-RELATED FEATURE EXTRACTCO EXTRACT TPR mm CALCULATE TIRST CALCULATE SECOND CALCULATE FOURTIT CHANGE VALUE CHANGE VALUE CHANGE VALUE RELATIVE TO TIME. OF CALIBRATION RELATIVE T0 TIME OF CALIBRATION RELATIVE 10 TIME ‘OF CALIBRATION ACQUIRE OFFSET BASED ON FOURTH TANGE VAL VALUE, AND OFFSETPatent Application Publication Feb. 20, 2020 Sheet 12 of 13 US 2020/0054290 AI FIG. 8A 800 820:Patent Application Publication Feb. 20, 2020 Sheet 13 of 13 US 2020/0054290 AI FIG. 8B DISPLAY PULSE WAVE SENSOR Sila Leen eedUS 2020/0054290 AL APPARATUS AND METHOD FOR ESTIMATING BLOOD PRESSURE. (CROSS-REFERENCE TO RELATED APPLICATION(S) 10001] This application claims priority from Korean Pat- ‘ent Application No. 10-2018-0096674, filed on Aug. 20, 2018, inthe Korean Intellectual Property Ollie, the disclo~ sure of which is herein incorporated by reference in its envi, BACKGROUND, 1. Fiekd 10002] Apparatuses and methods consistent wth example ‘mbosliments relate to estimating blood pressure, and more particularly, ro estimating blood pressure based on relative Change of « cardiovascular featre 2, Description of Related Art [0003] Recently, active research hus been conducted on Internet technology (IT}-modical convergence technology, Which is combination of IT technology ad medical technology. due 19 the aging population strcture, rapidly growing medical expenses, and the shortage of professional medical service personnel, The monitoring of the health status of the human body is no Limited to the bosptal, but is expanding to the fiekd of mobile health care, which ‘monitors the health status of users moving in everyday lie, suchas home and office. Archetypal examples of bio-signals indicating. the individual's health status may include an ‘lecirveardiography (ECG) signal, « photoplethysmogram (PPG) signal, an electromyography (EMG) signal, and the like, Various bio-signal sensors are being developed to measure such signals in daly life, Particularly inthe case of PPG sensor, it s posible to estimate blood pressure of human body by analyzing pulse waveforms in which a ‘cardiovascular statis is reflocted SUMMARY [0004] According to an aspect of an example embodiment, there is provided an apparats fr estimating blood pressure, including: a bio-signal sensor configured wo measure bio-signa from a user; anda processor configured to extract 2 feature fromthe bio-signal at an extraction time, acquire ‘an offset bused on a relative change value of the feature ‘eximcted atthe extraction time, with respect to a reference value of the feature obiained at a time of calibration, and ‘estimate a blood pressure based on the relative change vale ‘of the extracted fexture andthe sequired off [0005] The bio-signal may include atleast one of « pho- loplethysmogrm (PPG) signal, an_electrcardiography (BCG) signal, an electromyography (EMG) signal, a seis- ‘mocardiogram (SCG) signal, and 2 ballistoeardiogram (BCG) signal [0006] "The processor may be further configured to extract ‘cardiac output (CO) and atl peripheral resistance (TPR) from the bio-signal, as the feature [0007] ‘The processor may be further configured 10 Ssoquite, from the bio-sgnal, characteristic points compris- ing at least one of heart rate information, 2 shape of Waveform of the bio-signal, an area under the waveform, time and amplitude of 2 maximum point of te bio-sgnal, Feb. 20, 2020 time and amplitude of'@ minimum point of te bio-signal, ‘and amplitude and time of each of constituent pulse waver Torms constitling the bio-signal, and extract the feature related toa pulse pressure hosed on the charaeterstic points {0008} ‘The processor may be farther configured to cele Jate a first change value by normalizing the extracted CO based on a reference CO obtained at the time of calibration and calculate @ second change value by normalizing the extracted TPR based on. refereace TPR obtained at the time of calibration, [0009] |The processor may be futher configured to calcu Jatea third ehange value hased on the first change value and the second change value and acquire the offset by inputting at least one of the First, the second, and the third change values into an offset estimation modal {0010} » The processor may be further configured to deter. ‘ne atleast one ofthe fist, the second, and the third change values to be input tothe offset estimation model according to a preset time criterion. [OOL1] The processor may be farther configured to caleu Tate a outh change Value based on the first change value and a change in a heart rate obtained from the bio-signal, input the fourth change value into an offset estimation ‘model nd aequire the offet based onan output rest oF the offset estimation mode {0012} "The processor may be farther configured to caleu- Tate thi ehange value based on the first change value and the second change value, and acquire the offset by combin- ing the third change value wit the output result ofthe offset estimation model. [0013] The processor may be further configured (0 apply fa weight tothe first change value and the second change value, combine the weighted frst and second ehange values with the acquired offet to obtain a combination result, and tstimate the blood pressure by applying a sealing factor #9 the combination result [0014] The scaling factor may be set based on atleast one ‘fa mean arterial pressure (MAP) at the time of calibration, a diastolic blood pressure (DBP) atthe time of calibration, ‘and a systolic blood pressure (SBP) at the time of ealibrac [0015] The processor may be further configured to inde pendently estimate the MAP, the DBP. and the SBP by fagjusting at Teast one of the weight, the sealing factor, the first change value, and the second change value [0016] The processor may be further configured to ex ‘mate an amount change in the blood pressure hased on the relative change valve of the feature and the offset, and testimate the blood pressure using the estimated amount of hinge and a reference blood pressure obtained st the time of calibration, [0017] According to an aspect of another example embodiment, there is provided & method of estimate blood pressure, the method including: measuring a bio-signal from ‘user, extracting a feature from the bio-signal at an extra: tion ime: determining » change ina vale ofthe feature tha ‘curs during time period between a calibration time ofthe bio-signal and the extraction time; aequiring sn offset based fm the change in the value ofthe feature; and estimating a blood pressire based on the change in the value of the Teaure and the oft [0018] Theexiracting the feature may inelude extracting ardiae outpat (CO) and a total peripheral resistance (TPR) from the bio-signal, asthe featureUS 2020/0054290 AL [0019] ‘The extracting the feature may include acquiring, from the bio-signal, characteristic pons comprising a least ‘one of heat rte information, shape of a wavelorm ofthe biovsignal, an area under the waveform of the bio-signal, time and amplitude of # masimum point of the bio-signal, time and amplitude of @ minimum poiat of the bio-signal, snd amplitude and time of each of constituent pulse waver forms constituting the biosignal, and extracting the feature related to pulse pressure based on the acquired character- iste points [0020] The determining the change in the value of the feature may include calculating @ fist change value by normalizing the extrcted CO based on a relerence CO ‘obtained atthe time of calibration and calewlating a second ‘change value by normalizing the eximeted TPR based on a reference TPR obtained atthe tine of calibration, [0021] The determining the change in the value of the Featuce further may include calculating a tied change value based on the fst change value and the second change value and the acquiring the ollset comprises acquiring the offset by inputting the frst, the second, and the thi change vais 0 an offset estimation model. [0022] The determining the change in the value of the feature may further inchude calculating a fourth change Value based onthe first change valve and a change in a heart rate obiained from the bio-signal, and the sequiring the ‘offset comprises inputting the fourth change value into an “offset estimation mee! and aequiing the offset based on an ‘up result of the offset estimation model, [0023] The determining the change in the valuo of the feature futher comprises calculating a third change value based on the fist chinge value and the second change value, and the acquiring the ollst comprises acquiring the offsct by ‘combining the third ehange value with the ontput result of the offset estimation model 10024] The estimating the blood pressure may include ‘applying a weight wo the fist change value and the second change value, combining the weighied first and second ‘change values withthe offset to obtain a combination result, ‘and applying a scaling factor to the combination result 10025] The estimating the blood pressure may include ‘estimating an amount of change in the blood pressure based ‘onthe change in the value of the feature and the oflset, and ‘estimating the blood pressure based on the amount of change tnd a reference blood pressure obtained at the ime of calibration. 10026] According to an aspect of another example ‘ombodliment, there is provided an apparatus for estimating blood pressure, the apparatus including: bio-signal mea surer configured to measure a bio-signal from a user; ‘communication interface configured wo receive a pulse pres sure ofthe user fom a pulse pressure measurement appa Talus: and a processor configured to extract a feature from the bio-sianal, aequee an offset based on a relative change value ofthe received pulse pressure relative to 2 relerence value of the pulse pressure a time of calibration, and ‘estimate blood pressure based on the relative change value ‘of the feature and the off, 10027] The processor may be futher configured to extract, as the feature, a eatdiae output (CO) and a total peripheral resistance (IPR) from the bio-signl, calculate relative ‘change values ofthe CO, the TPR, and the pulse pressure relative tothe time of calibration by normalizing the CO, the TPR, and the pulse pressure, respectively, based on a ref- Feb. 20, 2020 erence CO, a reference TPR, and a reference pulse pressure ‘whieh are obtained at the time of calibration. [0028] The processors configured to apply a weight othe ‘ange value of the CO aad the change Yalve ofthe TPR, combine the weighted changed values of the CO and the TPR with the offset to obtain a combination result, snd estimate the blood pressure by applying a scaling factor to the combination rest BRIEF DESCRIPTION OF THE DRAWINGS [0029], Theabove andor other aspects wll be more appa cat by desribing certain example embodiments, wilt r= Srence tothe accompanying drawing, in whic {0030} FIG. 1 isa block diagram ilstrating sn apparatss for estimating blood presure aconting tone example embodiment: [0031] FG. 2s block diggram ilsuating an aparates for estimating blood pressore according to anchor eagle embodiments {0032} FIGS. 3A, 3B, and 3C are rans for deserbing an oiset estimation mode [0033] FIGS. 44 an 4B are block dsgrams illstaing cxample embodiments of» configuration of the processor according tothe example embodiments of FIGS, Hand 2, (0034) FIGS. §Aand SP are graphs for deserbing feature [0035] FIG. 6 is » Mowchar illustating » method of stimating blood pressure according 0 one example embodiment {0036} FIG. 7 is a Mowohart ilustating « method of cstimating blood pressure according to another example embodiment; and {0037} FIGS. 8A and 8B ace dvgrums for describing @ \wewable device acoeling W one example embeximent DETAILED DESCRIPTION [0038] Example embodiments are described in greater ‘etl below with reference tothe accompanying deawings [0039] "Inthe following description, ike drawing reference numerals are used for like elements, even in dillrent drawings. The matters defined in the description, such as ‘detailed constrution and elements, are provided to assist in a comprehensive understanding of the example embodi- tents, However, it is apparent that the example embod iments ean be practiced without those specifically defined rtters. Also, Well-koown functions or constructions ae not described in detail since they would obscure the description with unnecessary deta [0040] As used herein, the singular forms are intended to Include the plural forms a wel, unless the context clearly Indicates otherwise. It will be further understood that the teams “comprises” andor “comprising,” or “includes” and or “including” when used in this description, specify the presonee of stad features, numbers, stops, operations, elements, components or combinations thereof, but do not preclude the presence or addition of one or more other Teatures, numbers, steps, operations, elements, components ‘or combinations thereof. In addition, terms such as “unit” ‘and “module” denote units that process at least one function or operation, and they may be implemented by using hard- ware, software, or a combination of hanlware and software. [004t] Expressions such as “at least one of” when pre coding alist of elements, modify the entire list of elementsUS 2020/0054290 AL and do not modify the individual elements ofthe ist. For ‘example, the expression, “at least one of, b, and." should be understood a including only a, only b, en c, both a and b, both a and c, both b and e, all of a. b, and c, or any Variations of the aforementioned examples. [0042] Hereinafter, example embodiments of an apparatus ‘and method for estimating blood pressure will be described in detail with reference to accompanying drawings [0043] FIG. 1 iss block diagram illustrating an apparatas for estimating blood pressure aeconding to one embodiment. ‘The apparatus 100 for estimating blood pressure may be mounted in a terminal, such as a smartphone, a tablet personal computer (PC), a desktop PC, a notebook PC, or the like, er may be manufactured san independent are- ‘ware device. n this ease, the independent hardware device may be & wearable device, sich as wristwatch pe, & bracelet type, a wrist band type, a ring type, a plases-type, ‘oF @ hairband type, which ean be wor on an object of ierest. However, the apparatus 100 is not limited t0 the above examples [0044] Referring to FIG. 1, the apparatus 100 for estimat- ing blood pressure may include a bio-signal messuree 110 ‘and a processor 120, [0045] The bio-signal measurer 110 may include one oF ore sensors andl measure various bio-sgnals from an ‘object of interest through the sensors. The Bisignal mea surer 110 muy measure a photoplethysmogram (PPG), an ‘lectroeardography (PCG), selsmocardiogram (SCG), an ‘eleetromyouraphy (EMG), a ballistocardiogram (BCG), and fs pulse pressure. Examples ofthe biosighal measurer 10 may include a spectrometer, an optical sensor, PPG sensor, ‘an ECG sensor, a SCG sensor, an EMG sensor, BCG. seasor, and a pulse pressure sensor. However, the bio-signal easter 110 isnot limited tothe above examples, [0046] ‘The processor 120 may contra the bio-signal mea- surer 110 when a request for estimating blood pressure is received from a user oF when a eriteon for estimating blood pressure is satisfied, The processor 120 may reccive a biovsigual from the bio-signal measuree 110 ancl estimate blood pressure based on the received bio-signal, [0047] ‘The processor 120 may extract features that may affect blood pressure from the bio-signal. Examples of the features may include cardiac outpat (CO) and total periph- ‘eral resistance (TPR). [0048] For example, the processor 120 may acquire one or more characteristic points from the bio-sgnal ancl extract feature for estimating blood pressure by combining the soquired characteristic points. In particular, the processor 120 inay extract a one-period signal from the bios ‘continuously measured for # predetermined period of time. In addition, « representative waveform may be acquited from the extracted one-period signal and the characteristic points may be aequired using the aequired represent ‘waveform. In this case, a festure may be extracted by combining one or more pieces of time infonnation or mplitade information ofthe charscteriste points [0049] When the features extracted from the bio-signal at the time of extraction, the processor 120 may estimate blood pressure using the extracted feature and a reference feature ‘Obiained at the time of ealibration, For example, the peo= ‘cessor 120 may estimate a relative change of the extacted Featuze relative o a reference valve of the feature obtained atthe time of calibration (eg., a change in the extracted featured during atime period between the time of eaibration Feb. 20, 2020 and the time of extraction) and estimate blood pressure based on the relative change: The processor 120 may eal- culate «relative change vale of the feature by aommalizing the extracted feature based on the reference feature obtained atthe Gime of calibration and estimate blood pressure using the calculated relative change value, [0050] Meanwhile, since the CO and the TPR extracted trom the biosignal for estimating blood pressure ate ditfer- eat from the aetual CO and the aotual TPR, the processor 120 may acquire an offset for correcting such a difference and reflec the acquired offset to the blood pressure estima {0OSt] tn one example, the processor 120 may acquire the oflct by obtaining s fist change value by normalizing the extracted CO based on a reference CO abiained at the time of calibration and oblaining a second change value by formalizing the extracted TPR based on a reference TPR ‘obiained atthe time of calibration. For example, a thd change value may be calenlated by multiplying the fst ‘hange value by the second change value andthe offset may be acquired by inputting the calculated thi change value to fn offset estimation model. [00S2} In another example, the processor 120 may acquire ‘feature related to pulse pressure other than the CO and the ‘TPR and acquire the ose based on te feature related tothe pulse presue. The procestor 120 may acquire a fourth ‘change value based on a value obtained by normalizing the eature related to the pulse pressure based on a feature related to pulse pressure obtained atthe time of calibration fd acquire the offset by inputing the fourth change value to the olfiet estimation model [0083] In addition, the processor 120 may acquire, as a ‘ature related to pulse pressure, an actual pulse pressure measured from the user throogh the pulse pressure sensor ‘included in the bio-signal measurer 110 oa exteaal pulse pressure estimation apparatus. Also, the processor 120 may extract value corresponding to the pulse pressure ora heart rate asthe festure related tothe pulse pressure through pulse ‘eansit time (PTT), hear rate variability (HRV), peak ampli tude analysis, or the ike fiom the bio-signal [0054] The processor 120 may extra a CO feature and a "TPR feature for each blood pressure from the bio-signal, quire the olset for each blood pressure, or define & blo pressure estimation model for each blood pressure, thereby ‘estimating each blood pressure independently [0058] FIG. 2 isa block diagram illustrating an apparatus or estimating blood pressure acconting to another example eashodineat, [0056] Referring to FIG. 2, the apparatus 200 for estima ing blood pressure according to the present embodiment ‘may inehude a bio-signal measurer 110, a processor 120, ‘communication interface 210, an output interface 220, and a storage 230 {0057} The bio-signal measurer 110 may be electrically sonnected to the processor 120 and measure various bio- Signals from a user under the eontrol ofthe processor 120. [0058] The processor 120 may receive a bio-signal from the bio-signal measurer 110 and extract, for example, a CO feature and a TPR feature for blood pressure estimation ‘using the received bio-signal [0059] Acconding to the present embodiment, the proces sor 120 may aequire additional information necessary for blood pressive estimation from an extemal device 250, in addition to the features extrcted from the bio-signalUS 2020/0054290 AL Examples ofthe extemal deviee 250 may include an external blood pressure measurement apparatus, an external pulse pressure apparatus, and ober information processing appa fatuses, such as a'smartphone, a tablet personal computer (PO), a desktop PC, 4 notebook PC, und the like 10060] In one example, when the procestor 120 receives & rexuest for estimating blood pressure and a user is Wearing ‘or carrying an extemal pulse pressure measurement appa- ratus, the processor 120 may control the communication interface 210 to obtain a feature related to pulse pressure necessary for acquiring an offset. The communication inter- face 210 may communicate wih the external pulse pressure measurement apparatus under the control of the processor 120 and direlly receive pulse pressure information neces- sary Tor acquiring the offset fom the extemal pulse pressure measurement apparatus. The request for estimating blood pressure may occur when there is & user's iapit oF a predetermined eiteron is satisfied. In this case, the rede termined criterion may be a predetermined! measurement interval of a measurement time [0061] In another example, when the processor 120 receives a calibration request, the processor 120 may control the communication interface 210 to aequie reference infor- mation atthe time of calibration, The communication inter- face 210 may be connected to an external blood pressure measurement apparatus and/or the extemal pulse pressure Ieasurement apparatus under the control of the processor 120 and receive a reference blood pressure andor a refer- ‘ence ple pressure from the extertal blood presture © ration apparatos and/or the external pulse pressure mea surement appacats, The calibration request may occur when there is a users input ora calibration criterion is satisfied Here, the calibration erterion may be set in advance as a predetermined interval, the foal number of times when an ‘estimated blood pressure deviates from a normal range, the umber of consecutive times when the estimated biood pressure deviates from the normal ring, othe like [0062] The communication interface 210 may access a ‘communication network using a wiredvireless communi cation technology under the control of the processor 120 ‘The communication interface 210 may transit a processing result of the processor 120 to an external device 250, Examples of the communication technology may’ include Bluetooth communication, Bluetooth low energy (BLE) ‘communication, near field communication (NEC), wireless Tocal area network (WLAN) communication, ZigBee com- munication, infrared data association (IsDA) communica tion, WEFT diet (WFD) communication, wa-wideband (UWB) communication, Ant+ communication, WiFi com munication, dio frequency identification (RFID) commu- nication, 3rd generation (3G) communication, 4G commu- nication, SG communication, ete. However, the ‘communication technology is not Timited to the above ‘examples. 10063] | Meanvhile, the processor 120 may guide the user through the output interface 220 to measure blood pressure ‘andlor pulse pressure using the extemal device 280 in response to the calibration request oF the request fore rating blood pressure. The output interface 220 may output information for guiding the measurement of reference blood pressure andlor the measurement of pulse prestore using & Visual output module, such as a display. a voice output rmodile, sch as a speaker, ora apt madule using vibra- tion, tactile sensation, or the lik, under the contro of the Feb. 20, 2020 processor 120. The processor 120 may ontput a use inter Tice through the output interface 220. The user may use the wer interface to input blood pressure andor the pulse pressure measured through the extemal device 280 {0064} The processor 120 may store, as reference infor: ‘mation, a bio-signal measured through the bio-signal mea surer 110 atthe time of calibration a feature exircted from the bio-signal, the blood pressure andor the pulse pressure information measured through the external device 280 inthe Storage 230 for blood pressure estimation. In addition, the storage 230 may store & variety of other reference informa tion, such as user characteristic information, such as user's age, sex, health condition, and the like, a blood pressure fextimation model, an offset estimation model, andthe like [0065] Examples of the storage 230 may incude a storage ‘dium, such asa memory of flash memery type, band disk type, multimedia card micro type, or card type (eg. SD oF XD memory), random seoess memory (RAM), static ran- dom aecess memory (SRAM), read-only memory (ROM), lectrically erasable programmable read-only memory (BE PRON), programmable read-only memory (PROM), mag- relic meniory, magnctic disk, optical disk, or the ike, but is fot limited thereto [0066] The processor 120 may extract a feature from the bio-signal measured by the bio-signal measurer 110 in response to the request for estimating blood pressure and festmate blood pressure based on the reference information ‘obtained at the time of calibration, which is sored in the storage 230, In particular, when the user's pulse pressure information is measured though the external pulse pressare ‘measurement apparatus, the processor 120 may receive the ressured pulse pressure information through the comm ication interface 210 othe interface of te output interface 220 and acquire an offset based on the received usee's pulse pressure. The processor 120 may correct am eror between @ Co feature and a TPR feature, which are extracted from the bio-sgnal, aad an getual CO and an actual TPR, wsing the quired offset [0067] _ When the processor 120 competes the blood pres Sure estimation, the output interface 220 may output the bio-signal, the blood pressure estimation result, and ada ‘ional information ia accordance with the blocd pressure estimation resul. In one example, the output interface 220 ‘may provide a variety of information visually through Aisplay module, For example, when estimated blood pres- sure deviates from a normal range, the blood pressure result ‘may be emphasized in red color, thereby presenting warning information to the user In ancther example, a variety of information may be provided tothe user through a speaker ‘or haptic modile in a non-visual manner, such 38 a voice, vibration lacie sensation, or the like, For example, systolic blood pressure (SBP) and diastolic blood presture (DBP) ‘may’ be informed by voice, When estimated blood pressre deviates from a predetermined normal range, the weer my be informed of abnormality in health condition throwgh vibration oF tactile sensation. [0068] | When the processor 120 competes the blood pres- sure estimation, the storage 280 may store the bio-signal andlor the blood pressure estimation result, [0069] | FIGS. 34.0 3C are grap for describing an offset cestimation model [0070] Generally, an amount of change ia mean axetil pressure (MAP) is proportional to the CO and the TPR as Shown in Equation T belo.US 2020/0054290 AL [0071] Here, AMAP denotes a diference in MAP between the left ventricle and the right ventricle. Generally, mean right ventricular pressure does not exceed 3 to 5 mums and is similar © mean left ventricular pressure or brah mean arterial pressure. Thus, when absolute CO and TPR values ‘re knoWn, its possible to aban the MAP in the aorta or ‘anupper arm. However, itis difficult to estimate the absolute ‘CO and TPR values based on the bio-signal 10072] In general, values within a range of plus or minus 0510027 ofthe MAP from the MAP calculated as described ‘sbove may be used as SBP and DBP. Llowever, a decoupling Phenomenon may occur in which SAP and DBP do not follow the tendency of change of MAP according 10 mechanism of blood pressure change, In addition, ina case of, for example, a igh intensity aerobic exercise, in which the CO or the TPR clianges significantly relative to a stable ‘condition, an error of estimated blood pressure may be ‘greatly increased. 10073] For example, as shown in FIG. 3A, in various reclsnists of blood pressure change, the MAP may remain the same while DBP and SBP may change. This indicates a ‘case in which the DBP and the SBP change according tothe ‘change in pulse pressure even sthen the MAP does not change. 10074] According to the present embodiment, blood pres- sure may be estimated based on relative changes i CO and “TPR features relative tothe time of calibration, Since the pulse pressure may be determined based on changes in Snerial stiffs and stroke volume (SV), an offet for ‘correcting the exror due to the amount of change in pulse pressure may be acquired using Features related to the Aanerialsifness and stroke volume and the obtained offset may’ be used in blood pressure estimation, thereby impro¥ ing the accuracy. 10075] FIGS. 3B and 3C are prophs for deserbing an ‘offset estimation model, The offset estimation model may be in the form of a quadratic funtion as shown in Equation 2 below, However, the oft estimation model is not limited thereto, and may be a linear finetion, a piscewise linear finetion, a mult-dimeasion nonlinea Function, oe the like rarer nie @ 10076] Here, X denotes an input value and V(X) denotes ‘an offSet for input X. In adiion, a, by and ¢ are predeter- nined constants, which are obtained through preprocessing as described with reference to FIGS. 38. and 3 That is, input X of the offset estimation model may be defined ‘variously as one of or a combination of two or more change values describod above, 10077] FIG. 38 shows an example in which an offset ‘estimation model for acquiring an affsct is derived using 3 third change value AfTxA(2 calculated based on a fist ‘change value Af] and a second change value AB. The first, ‘change value MI-may be oblained by normalizing an cexineted CO by a reference CO, and the seeond change value A12 may be oblained by normalizing an extracted TPR by a reference TPR. The offset estimation model may be derived by acquiring coefiiens a, band ¢ thereof using a third change value AMxAD and an optimal offst value ABP-(Afl+A12) of each sample ata plurality of measure- ment times. Here, ABP represents the actual amount of change in blood pressure. Feb. 20, 2020 [0078] FIG. 3C is a graph showing an example in which ‘an ob estimation mode for acquiring an oft is decived based on a fourth change value related to pulse pressure Coeficiente a,b, and e of an offset estimation model are aequired using & fourth change value ASV/AHR and an optimal offset value ABP-(Af1+A12) at each point in time ante offset estimation model may be derived, wherein the Tourih change value ATI/ATIR is calculated based on a vale AHR obtained by normalizing heart rates of each sample obtained at @ plurality of measurement times and fst change value Af [0079] FIGS. 44 and 4B ae block diagrams illsiating configuration of the processor according to the embodiments of FIGS. 1 and 2. FIG. 5A is a diagram for describing eximetion of a feature for blood pressure estimation, [Embodiments of blood pressure estimation performed by the processors 410 and 420 will be described with reference to FIGS. 44 and 4B. [080] Referring 10 FIG. 4A, the processor 410 may Include a feature extractor 411 a feature change estimator 4412, an offset acquirer 413, and a blood pressure estimator aa [OSE] The feature extractor 411 may extract a cardiovas ‘ular festure from various bi-signals measured from a use. In this ease, the cardiovascular Teature may include a CO ‘ature and a TPR feate [0082] For example, the eature extractor 411 may acquire heartbeat information, a shape of a waveform of & bio- signal, time and amplivde of @ maximum point of the bio-gnal, time and amplitude of @ minimum point of the bio-signal the area of the bio-signal waveform, elapsed time of the biosignsl, amplitude and time information of each constituent pulse of the bio-signal from the bio-signal, aequire characteristic point information, suet 3s information on an internally dividing point between the pieces of foquired information, and extract a feature using the acquired characteristic point information [0083] FIG. 5A shows an example of a pulse wave signal ‘among the bio-signals acquired from the user. One example in which the featire extractor 4 extracts feature from the pilse wave signal will be desribed with reference to FIG. 5A, [0084] In genom, pulse wave signal is a superposition of ‘8 propagation wave propagating from the heat to vascular bifurcations ofa body and reflection waves returing from the vascular bifurcations. FIG. SA illustrates that a wave- orm of measured pulse wave signal PS is a superposition of five constituent pulses, for example, a propagation wave fiw and reflection waves rw, rW2, rW3, and Fwd [0085] The feature extractor 411 may acquire charactris- tie points by analyzing wavefoms ofthe constituent pulses fix ew, n02, 3, an rw from the pulse wave sigoal PS. The first three constituent pulses fs, R01, ad r02, may be used to estimate blood pressure. Subsequent pulses may not be observed in some users, may be difiult wo devet due to ‘oise. oF may often have low cortelatinn with blood pressire estimation, [0086] For example, times T,, T, and T, and amplitudes P,, Pa, and Ps of maximum points of the frst to thd constituent pulse wavefomns fv, rv, and ew2 may be ‘obtained as characteristic points, When 8 pile wave signal PS is obiained, a second-order derivative of the obtained pulse wave signal PS is computed andthe times T,. Ts, and 1, and amplitudes P,,P., and P, of maximum points oF theUS 2020/0054290 AL ‘constituent pulse waveforms fw, rw1, and rw2 may be ‘obtained using the obtained second-order derivative signal. For example, local minimum points are scarce for in the scoond-orer derivative signal to extact times TT, andT, ‘coresponding to the fist o thi local miniznua points and the amplitudes P,,P,, and P, comesponding to the extracted times'T,, Ts, and, may be extracted from the pulse wave signal PS, Flere, the local minimum point refers oa specific Point observed ina segment ofthe sscond-onder derivative ‘anal, at which the signal stops falling and stants to rise ‘agua. That is, the local minimum point refers toa downward conver point. However, the embodiment is not limited thereto, sel that loca maximum points may be searched for in the Second-order derivative signal to extract times and amplitudes. In tis ase, the loa maximum point refers 9 8 specific point, observed ina segment ofthe second-order lerivative signa at which the signal eases to rise and starts to fll again, That i, the local maximam point refers to an upsvaed convex point [0087] In another example, de feature extractor 411 muy ‘bia time Ty a amplitode Py a @ pot ina pred termined iter ofthe pulse wae Signal PS at which the tmpltde is maximum as the characteristic pats. The Protermined interval may refer to an interval from the beginning ofthe pise wave signal 10 @ point where the Alico notch (DN) occurs, whieh indicates blood pressure syste period 10085] In another example, de feature extractor 411 nny ‘hinin time dumtion PPC, indicating the total mcasine- tuent te of the biosigna or the area PPG, under the biowsignal waveform sth earacterstic pointe, thearea PPG... under the bo-sgnal waveform may refer to the tt area der the waveform of the bio-sianal oF the ‘rc under the wavelonn ofthe bi-signal corresponding 8 predeteminal proportion (eg, 70%) of the eae me “ration PPG [0089] In sill another example, the feature extractor 411 ray extract an internally dividing point berween two oF more characteristic points as an addtional characteristic Point, Unstable waveforms ofthe pulse wave signal may be ‘generated duc 19 non-ideal environment, such as motion noise, sleep, and the ike so that the characteristic points may’ be extractod from wrong positions. In this case, blood pressure measurement may be supplemented by utilizing an ‘ntermlly dividing poim between the erancously extracted charsteratie points 10090] For example, when charscteristie points (T,, P,) and (Tres Pras) afe obtained from the Blood pressure systolic period iis possible to obtain an intemally dividing Point (Ty, P..) between the wo characteristic points (T,, ,)and (Py Byga)-n this ease, Weights are applied to time values T, nd Too of the two characteristic points (T,. P,) ‘and TexPgo) tine T,, oF the interaally dividing point nay be obtained using the weighed time values, and an amplitude P,, corresponding to the time T,,, of the inter- nally dividing point may'be extracted, However, the embod ‘ment isnot limited thereto, such thal through the analysis of the obtained bio-signal waveform, an intemally dividing point between characteristic points (Ty, P,) and (TP) Felated tothe frst and second constituent pulse wavelorms fiw andr, may be obiained from the blood pressure systolic period and aa iaterally dividing point hetween character: Fate points (Ts, Ps) and (Ty Py) related to the thd and Feb. 20, 2020 fourth consistent pulse waveforms rv and rw fom the blood pressure diastole period may be obtained. [0091] The featur extractor 411 may extract a CO feature ad # TPR feature by combining the variows characteristic points obtained from the bi-signal 1s described above. For ‘example, features may’be extracted by performing multipi cation, division, addition, subteaction, or a. cormbination thereof on the plurality of characteristic points, Alter tively, features may he extracted sing fiction that wes a result of performing multiplication, division, addition, subtraction, oF a combination thereof on the plurality of characteristic points as an input value. Here, the function ‘may be a Tinea function, a quadriefuneton, another mult dimensional function, a logarithmic function, or an expo- ential funeton, Iti apparent that ather types of function «an be used. In another example, a feature may be extracted ting afnction that nses at least one characteristic point as ‘an input valve. However, the embodiment isnot ited to the above examples, [0092] Meanwhile, the CO feature and the TPR feature ray be extracted by combining the characteristic points diferently according t characteristies ofthe user. In ad ‘ion, the CO feature and the TPR Feature may be individually extracted foreach iype of blood pressive by combining the characteristic points according to the blood pressure to be extracted, for example, MBP, DBP, and SBP. [0093] In ation, the feature extractor 411 may extract a feature celated to pulse pressure from the measured bio- signal, For example, «heat rate (HR) may be extracted a feature related to pulse pressure through an analysis of pulse transit ime (PTT) heat rate variability (HRV), o the like of the measured bio-signal. In addition, as shown in FIG. 8B, estimated pulse pressure which is proportional 4K peak amplitude of seismocardiogram (SCG) signal ‘ra ballistocardiogram (BCG) signal may be extracted as a feature related to pulbe pressure [0094] When the feature for blood pressure estimation is fexinicted, the feature change estimator 412 may estimate a relative change in the extricted feature using a reference Teature obtained atthe time of calibration, Por example, the ‘ature change estimator 412 may calculate 9 fist change value by normalizing an amount of change in CO obtained at the time of blood pressure measurement relative 10 a reference CO obtained at the time of calibration to the reference CO. In addition, the Feature change estimator 412 may caleulate a second change value by normalizing an amount of change in TPR obtained at the time of blood pressure measurement relative t0 a reference TPR obtained ft the time of calibration tothe reference TPR, {0095} Also, for each additional feature for offset acqui- sitio, the feature change estimator 412 may calculate a relative change Value reatve to the Gime of calibration. In ‘ne example, the fenture change estimator 412 may calelate ‘a thied change value by multiplying the frst change value by the second change value. In another example, 2 fourth change value may be calculated by normalizing the feature related to pulse pressure oa feature related to pulse pressure ‘obtained at he time of calibration. In this ease, when the ature related to pulse pressure isan IR, te feature change estimator 412 may calculate @ HR change value by noemal- ining the extracted IIR toa reference HR obtained at the time of calibration and the ealeuate the fourth change value related to pulse pressure by dividing the first change value by the AR change valueUS 2020/0054290 AL [0096] The offset scquirer 413 may acquire an offset by inputting the change value calculated for offset acquisition to'an ollset estimation model, Ia one example, the offset soquiter 413 may acquire an offset By inputting the third change value to the offset estimation model. In another ‘example, one ofthe fist to thin! change Values is selected according to a predetermined criterion and the selected ‘change value may be input to the offset estimation mode For example, a tiresbold may be set in advance for each ‘change value, nd a change Value that exceeds the threshold ‘ora change vale that exceeds the threshold by the greatest mount may be determined to be input to the offset estima tion model. However, the embodiment is not Timited to the ‘ove examples. [0097] In another example, an offet may be acquired by ‘nating the fourth change valu related to pulse presse to the offset estimation mode. In another example, the fourth ‘change value related to pulse pressure may be input w the ‘offset estimation model and final offset may’be acquired by. ‘combining an output result of the offset estimation model with one of the fist to third change values. [0098] The offset estimation model may be defined in advance in consideration of the computing performance of the apparatus 100 for estimating blood pressure. type ofthe bio-signal measurement sensor mourted in the apparatus 100, and various measurement conditions, such as user's characteristics. An input of the offset estimation model may be determined according to the offset estimation model applied to the apparatus 100. [0009] ‘The blood pressure estimator 414 may estimate an ‘amount of ehange in blood pressure by inputting the frst, ‘change value and the socond change value estimated by the feature change estimator 412 and the eset acquired by the “offset acquirer 413 to a blood pressure change estimation model. For example, one example of the blood pressure ‘change estimation modal is shown as Equation 3. Referring to Equation 3, the amount of change in blood pressure may be calculated by a linear combination of the fist change value and the sevond change value with the offset, In this ‘ase, weights may be assigned tothe first change value and the second change value and a scaling factor may be applied to a result of linear combination, so that the amount of ‘change in blood pressure which reflects a characteristic of ‘each stscr may be obtained ABP srxaanenanvomen © 0100] Here, ABP denotes sn estimated amount of change in blood pressure, and may be MAP. DBP, or SBP. AfT and ‘M2 denote the first change value and the second change value, respectively. Offset denotes an offiet acquired by the ollket aoquier 413 using the oflset estimation model shown, in Equation 2. The variables «t and fi denote a weight ‘ssigned 1 each change value and may be defined according toa ‘pe of blood pressure to be estimated andior the characteristics of the User. Inston, SF denotes a scaling factor which i defined adaptively according to the eharac- teristics of the user andor the type of blood pressure 10 be ‘estimated, For example, the sealing fctor may’ be reference MAP, reference DBP, or reference SBP, whieh is measured fiom the user by an extemal blood pressure measurement ‘apparatus a the time of calibration, oF value calculated by ‘combining two or more ofthe reference MAP, the reference DBP, and the reference SBP Feb. 20, 2020 [0101] Meanie, the blood pressure estimator 414 may independently estimate the amount of change in MAP, DBP, ted SBP using Equation 3 above. For example, a feature for estimating blood pressure may be extracted for each type of blood pressure al the amount of change in blood pressure of each type may be independent estimate. Alternatively, weight and/or a sealing factor may be set differently for och type of blood pressure and the amount of change in blood pressure of cack type may be independently esti- mated. For example, reference MAP of the user may be used as a sealing factor to estimate MAP. Similarly, reference DBP and reference SBP may be used as sealing factors #9 estimate DBP and SBP. [0102] When the amount of change in blood pressure is estimated, the blood pressure estimator 414 may estimate blood pressure using a blood pressure estimation model as shown in Equation 4, Here, the blood pressare estimation ‘model is illustrated as linear combination, but is not Timited thereto. [0103] Here, BP.., denotes estimated blood pressure, and ABP denotes” an ‘xtimated amount of change in blood pressure. In addition, BP,,, denotes reference blood pressure atthe ime of calibration For example, estimated MAP may be ealelated by inpatting en amount of change in MAP and reference MAP to the blood pressure estimation model Similarly, SBP and DBP are estimated in the same manner as the MAP, so that each type of blood pressure ean be independently estimated [0104] In another example, the blood pressure estimator 414 may sequentially estimate MAP, DBP, and SBP. For example, the blood pressure estimator 414 may estimate MAP through the above Equations 3 and 4 and estimate DBP and SBP using the estimated MAP and polse pressure For example, when the feature related to pulse pressure is foquired as described above, the blood pressure estimator 4414 may estimate DBP and SBP through Bquations 5 and 6 below, which are examples of functions for estimating DEP and SBP, pers sar ° [0105] Here, MAP denotes estimated mean arterial pres sure, DBP denotes diastolic blood pressure, and SBP denotes ‘solic blood pressure, In addition, PP denotes pulse pres- sre, and HR denotes a heat rate [0106] Referring t0 FIG. 4B, the processor 420 may lrer include @ calibrator 418 in addition to the cong ration ofthe processor 410 shown in FIG. 44. [0107] When a calibration request of the user is received, for when whether calibration is ruined is determined by referring to preset calibration criteria as described above and. the calibration is determine to be required, the calibrator 4415 may perform calibration, {0108} When performing calibration, the calibrator 418 may provide guidance for calibration 10 the user. For cexample, the calibrator 418 may guide the user to be inUS 2020/0054290 AL contact with the biosignsl measurer 110 to measure a reference bio-signal In alton, the calibrator 18 may provide auidence about a Gime point of measurement of Blood pressure, « measurement method, a communication ‘oanection beeen an extemal lood presstee messin tent operas and the communication iterfoce 310 0 hat. the user measure the reference Blood pressure though the ‘extemal Blood pressure messirement apparats. In adion, ww the users the ple pressure measurcncal apart Separately the calibrator 418 may provide guidance about time point of measurement of pulse pressure, a eommuni- ation coanection, an the ike to measure polse pressure throu the external pulse pressure measurement apparats [0109] When the reference blood pressure andor refer. ‘ence puke presse are measure thot the exter blood Prost meosorcment sppomtis andor the exteral poke pressure measurement appara, the calibrator 18 may ‘quire the reference blood pressure andor the reference Pulse presure frm the extemal blood pressure mcasir= ment aipparats andor the external puke presse measur tent apparats. La parca, the refereace Dowd presse tnd the reference ple presire may he rcsived dieetly throug the communication eonnection with dhe eommuni- ‘ition interface 210 or from the usr through 8 Wer ine face {o110) When referee features, such as CO, TPR, a feature related to pulse presse andthe lke ate extacied by the feature extractor 410, te calibrator 18 muy receive the reference features frm the feats extractor 410, Jolt] Tye calibrator 418 auy sore the syuire reference blood pressire, reference pulse presse, reference featres in the storage 230 as ceereace information for blood pes: [0112] The calibrator 418 may calate an ofiet estima tion mod! ade a blood peesure estimation mod hated fon therefore infomation. Foe example, the eaibrator 4418 may acquize the information, suchas reference bier signi, the reference blood pressure, the reference pulse presse, the reference festres, andthe ike, ata pray a Points in ine fora predetermined period ofime and aeyhire 5 acw coclicient of the offct estimation mod! using the method deseribed with reference fo FIGS. 3C and. 3D, thereby caliating the offset estimation model, Altea tively the calbstor 418 may calibrate the blood presire ‘estimation model by applying newiy acquired reference blood presse for cabration tothe blo presse estima: tion mods as shown in Equation 4 0113} FIG. 6 is a flowchart illostatng «method of ‘timating blood presue according tone example ‘mbit. [0114] FIG, 63s one embodiment of a method of estinat- tng blood pressure in accordance with the embodiment of FIG. 1 or 2. Various embodiments are described in desi shove, and hereinafter he method willbe describe in be. 0115} First, an apparatus 100/200 for estimating blood pressure may ceive a request for estimating blood presse ‘operation 610, The spparats 1002200 may provide an inverfce to a user and receive the request for estimating blood pressure input by the user dyough the interface Altematively. the appara 100/200 may comminicate with fan external device and receive the reuest foe estisting blood presse from the extemal deviee. In thin cae, the ‘external device may bes smariphone or a tablet PC erred Feb. 20, 2020 by the user and the user may control an operation of the apparatus 100/200 for estimating blood pressure through the tetera device [0116] Then, the apparatus 100/200 for estimating blood pressure may acquire bio-signal from the user by conto Ting a sensor (ee, a bio-signal measurer 110) intemally ‘mounted for blood pressure estimation or receive 0 ios signal from an extemal sensor in operation 620. The sensor ‘mounted in the apparates 100'200 and the extemal sensor ‘may acquire various bio-signals, such as @ PPG signal, an ECG signal, an BMG signal, an SCG signal, and a BCG signal, from various body pats (e., west, chest, fingers, te.) of the user. [0117] Then, the apparatus 100/200 may extract a CO feature and a TPR feature by analyzing the aeguired bio- signal i operations 631 and 632, Ia particular the apparats 1001200 may acquire, form the bio-sgna, characteristic points including information, sueh as heat rate information, f shape ofa waveform of the bio-signl, time and amplitude of a maximum point ofthe bio-signal, time and amplitude of ‘minimum point ofthe bio-signal, the area of the bio-signal ‘waveform, elapsed time of the bio-signal, and amplide and time of each coastinint pulse wavelom of the bio-signal, ‘and an interally dividing point between two or more characteristic points, and extract a cardiovascular feature by ‘combining the pies of acquired information. [O118] _A first change value that sa relative change ofthe CO feature relative to the time of calibration may’ be calculated in operation 641 and a second change valve that is a relative change of the TPR feature relative tothe time of calibration may be calculated ia operation 642, For example, a relative amount of change in CO feature relative to reference CO obiained atthe time of calibration may be normalized wo the reference CO to ealeuate the first change vale. Similarly, a relative amount of change in TPR Teatare relative to reference TPR obtained atthe time of ealibvation ‘ay be sonnalized to calculate the second change value [0119] A thind change value may be ealeuated based on the frst change value and the second change value in operation 648, [0120] An offset may be acquired by inputting the aleu- lated third change vale to the offset estimation model in ‘operation 650. However, the embodiment is not limited thereto such that a change value satisfying predetermined criteria among the fist change value, the second change value, and the thitd change valve may’ be input tothe offset estimation model. [0121] Thea, the apparatus 100/200 for estimating blood pressure may estimate blood pressure based on the fist thange value, the second change valve, and the offset in ‘operation 680, When blood pressure is estimated, the appa rats for estimating blood pressure may provide a blood pressure estimation reslt tothe user. [0122] FG. 7 is @ flowchart ithastating another embod ‘ment of the method of estimating blood pressure in accor- dance with the embodiment of FIG. 1 or 2. Tae method is described in detail above, and hence wil he described below in brief [0123] First, the apparatus 100/200 for estimating blood pressure may receive & request for estimating blood pressure fm operation 710 and acquire a bio-signal of user in ‘perntion 720, [0124] A CO feature may be extracted by analyzing the ‘oquired bi-signal in operation 731 and a TPR feature mayUS 2020/0054290 AL also be extracted in operation 732. In ths ease, one or more ‘characteristic points ae acquired from the bio-sigaal andthe (CO feature and the TPR feature may be extracted by ‘combining the acquired characteristic points. In this case, ‘each feature for estimating MAP, DBP, and SBP may be ‘exinicied by differently combining the characteristic points, In addition, a feature related pulse pressure maybe ‘exineted for aequiring an offset in operation 733. In this, ‘eas, the feature related to pulse pressure may include 3 hear rate 10125] fist change value that is a relative change of the CO feanire relative to the time of calibration may be ‘calculated in operation 741 and a second change value tht, is a relative change of the TPR feature relative tothe time ‘oF ealibration may be culeulated ia operation 724, In addi tion, @ fourth change value may be caleulated based on a relative change value of the pulse pressureelated feature relative 10 the time of calibration in operation 748, For ‘example, when a heart mite is extracted asa relative change value ofthe pulse presure-elated feature, the fourth change Vale that correspond to pulse pressure may be calculated by dividing the first change value by the relative change value of the heat rate, However, the embodiment is not limited tothe above example, and when actual pulse pres- sure measured through an extemal pulse pressure measure ment apparatus is input, the fourth change value may be ‘calculated by normalizing the actual pulse pressure based on the pulse pressure obtained atthe time of calibration 0126] An offset may be aequired by inputting the ealen- Tated fourth change value to a offset estimation model in ‘operation 750, In this case, a thint change value may be ‘calculated based on the first change value and the second ‘change value, and a final offset may be obianed by linearly ‘combining the calculated thi change valve with an est mation result ofthe offset estimation model [0127] Thea, the apparatus 100/200 for estimating blood pressure may estimate blood pressure hased on the fist ‘change value, the second change value, and the offset in ‘operation 760, [0128] FIGS. 8A and 8B are diagrams for deseribing ‘wearable device according to one embodiment. The above- deseribod various embodiments ofthe apparats 100°200 for ‘estimating blood pressure may be mounted in a smart wate ‘wom on a wrist or a smart band type wearable device as shown in FIG. 8A. However, the embodiments are not limited thereto, and the apparatus 100/200 my be mounted ina device, such as a smariphone,a tablet PC, a desktop PC, ‘8 notebook PC, and the like. [0129] Referring to FIGS. 8A and 8B, the wearable devioe ‘800 may include a main body 810 and a strap 820. 0130] The strap 820 may be formed tobe flexible s0 as 9 he hent ina shape to wrap around a wrist of a user oF tobe sachs from the user's wrist, Altematvely, the stp 820, may’be configured in the form ofan undivided band inthis, the strap 820 may be illed with airor have an air hag to have elasticity acconting to a change in pressure upplied to the ‘wrist and may transmit the pressure change of the wrist to the main body 810, [0131] A battery for supplying electric power to the wear- able device may be embedded inthe main body 810 or the snp 820, 10132] In addition, one or more sensors for measuring ‘various bio-signals may be mounted in the main body 810 ‘which is brought into contact with an object OB (eg, a Feb. 20, 2020 wis). For example, a pulse wave senor 811 may be ‘mounted on a rer stefac of tho main bly 810, in sich a ‘nr tat is expose tothe object OBJ. The pulse wave Scosor SIL my inch a light source 8a configured to ent Hight othe objet OBS anda detector 118 cntigued {© measure a pulbe wave signal by detecting ight sated or reflected from the object OBJ. In this ease, the light Source Bila may include at last one of o Baht emiting diode (LED), laser diode al a phosphor. and may be configured asa single aay’ or two or more aay [0133] "The min body 810 of the wearable device 800 Imay elude a processor 812 configured to estimate blow pressure based on the bio-signal recived from the pulse wave sensor 8H andor an exleral sensor The processor $12 may conto the pulse wave sensor 811 hy generating a onl signal in espouse to user's request for estima ‘ood prestie, and, when accessory. conto the comm cation interface B13 10 receive the bio-sgnal from the extemal sensor {0134} The communication interice 813 may be mounted fase the main body 810 and communicate with te exter device under the contol of the procssne #12 to transmit and receive necessary information. For example, the comm Cation interface 813 may recive the io-signal om an extemal sensor for measuring a bios for example, an FECG sensor, an EMG senso, a BCG senso, pulse prose sensor and the like. In adtioa, dhe communication intr {aco 813 may receive a request forestmating Blood press from the user's mobile terial. aditon, the eam cation intrfae 813 may cans the external device to estimate blood pressure by tansmiting the exacted har ater point feature information othe extemal deve ‘so, th Communication interlace 813 may transit blood pressure estimation ees to the external device sch hat the ‘ood pressure estimation rest ea be displayed to the ser o utilized for various purposes, such ab Blood pressure story management disease rescrch and he ike. Further, the communication interne B13 may resve a bood pre sue estimation equation oF reference information, such 38 reference blood prossire messrod by blood pressire ‘ewurement apparatus fom lhe extra device {0135} When the bi-sianal is reeived fom the pulse wave sensor #12 andor the extemal sensor the processor B12 may extract 2 CO feature and a TPR feature fom the received bio-signl For example, the above-described var- tu characteristic points may be acquired by analy the pulse wave signal and the festire may be extated by Combining the aequied characteristic points. In this ese, tho processor 812 may extract tho CO fete and the TPR feature for each of MAP, SBP, and DBP. [9136] The processor 812 may estate relative changes in CO festure and TPR festre ela to the ime of calibre tion, taking ino consideration difiulty in obiaining abso Jute values of the CO eatire and the TPR feat, nd cotimate blood pressure using. the estimated relative changes, In this ase, in ord inereake the accuracy of ood pressure estimation the blond pressure may be et- sated by rollecting a change in pulse pressor that alles te blood pressure (0137) For example, the procesor 812 may sequie an ost that represents an effsct de tothe change sn alse pressure based on the relative changes in CO Teatee and {TPR feature and apply the scquired offset 10 the blow pressure estimation, Alematvely, the procesor 812 myUS 2020/0054290 AL ‘extract a feature related to pulse pressure based on the pulse ‘wave signal measured through the pulse wave sensor 8 or the signal received from the external sensor and aeguie the ‘offset using relative change in the extracted feature related to pulse pressure relative tothe time of calibration. In this case, the feature related to pulse pressure may include information about heart rate'or actual pulse pressure 10138] The wearable device 800 may furher include an ‘operator S18 and a display 814, which are mounted in the main body 810, [0139] ‘The operator $15 may receive a control command from the user and deliver the control command 10 the processor 812 and include s power bution used for inputting 5 command for taming anf the wearable device 800 [0140] The display 814 may provide a variety of informa- tion related to the detected blood pressure to the user under the contol ofthe processor 812, For example, the display 814 may provide addtional information, such as detected blood pressure, alarm, warning, and the like, to the wer in ' Visual/aoa-visual manner, [0141] While not restricted thereto, an example embol- ‘ment can be embodied as computerreadable code om & ‘computerreadable recording mesium. The computer-rea able recording medium is any data stomge device that ean ‘tore data that can be thereafter read by a computer system. Examples of the computer-readuble recording medium inchide reachonly memory (ROM), rixdom-access memory (RAM), CD-ROMS, magnetic tapes, lopry disks, and opt ‘al dala storage devices. The computer readable recording medium can also he distributed over network-coupled com puter systems so tht the computer-eadable code is stored ‘and exeetted in a distbuted fashion. Also, an example ‘embodiment may be written as a computer program trans= mitted over a computer-readable transmission medium, such ae carrer wave, and roveived and implemented in gneral- tise or special-purpose digital computers thot execute the programs Moreover, it is understood that in example ‘embodiments, one or more units of the above-described apparatuses and devices can include eircuity, a processor, & microprocessor, et, and may execute a computer program Stored in a computer-readable medium, [0142] ‘The foregoing exemplary embodiments are merely ‘exemplary and are not 10 be construed a limiting. The present teaching can be readily applied to other iypes of apparatuses. Also, the description of the exemplary embod ries i intended vo be illustrative, and not it the scope of the claims, and many allematives, modifications, and ‘variations will be apparent to those skilled in thea. ‘What is claimed is: 1. An apparatus for estimating blood pressure, the appa- ratus comprising: 1 bio-signal sensor configured to measure a bio-signal of fiser and processor configured 10 extract a feature from the bio-signal at an extraction time, acquire an offset based ‘ma relative change value of the feature extracted at the extraction time, with respect oa reference valve ofthe {eature obtained at atime of calibration, and estimate 3 blood pressure based onthe relative change value of the extracted feature and the acquired ost 2, The apparatus of claim 1, wherein the bio-signal ‘comprises at least one of a photoplethysmogram (PPG) signal, an electrocardiography (BCG) signal, an electeomyo~ Feb. 20, 2020 raphy (EMG) signal, a ssismocardiogram (SCG) signal, And a balistocariogram (BCG) signal 3. The apparatus of claim 1, wherein the processor is turer configured to extract a cardiae output (CO) and a total peripheral resistance (TPR) from the bio-signal, a the reature 4, The apparats of elaim 3, wherein the processor is rer configured to acquire, fom the bio-signal, ehurac- ‘erste points comprising st last one of bear mate informa. tion, a shape of a Wavelorm ofthe bio-signal, an area under the Waveform, dime and amplitude of a maximum point of the bio-sgna, time and amplitude ofa minimum point ofthe bio-signal, and amplitude and time of exch of constituent pulse waveforms constituting the bio-signal and extract the Teature hased on the characteristic points, 8. The apparatus of elaim 3, wherein the processor is furter configured to calculate a first change value by normalizing the extracted CO based on a relerence CO ‘obtained atthe time of calibration and calculate a second change value by normalizing the extracted TPR based on reference TPR obtained atthe time af calibration 6. The appartus of elsim §, wherein the processor is Jrtuer conigured to ealculate a third change valve based on the first change valve and the second change Value and acquire the offset by inputting atleast one of the first the second, and the third change values into an offset estimation model 7. The apparatus of claim 6, wherein the processor is further configured to determine at least one ofthe fist, the second, and the thint change values to be input othe offset estimation model according to a preset time criterion, 8. The apparatus of claim 8, wherein the feature futher includes feature related to pulse pressure 9. The apparatus of elaim 8, wherein the processor is rther configured to calelate a fourth change value by sormalizing the feature related to pulse pressure based on a feature related to pulbe pressure oblained at the time of calibration, input the calelated fourth change valle ito an bffsct estimation model, and sequire the offset hased on an output result of the offset estimation model 10. The apparatus of claim 9, wherein the processor is rer conighred o calculate a third change vale bese on the first change valve and the second change valve, and acquire the offset by combining the thin change value with the output result of the offset estimation model 11, The apparatus of claim 8, wherein the processor is Jurher configured to apply a weight tothe fist change value and te second change value, combine the weighted fist and second change valves with the acquired olltet to obtain a combination result and estimate the Blood presse by applying a sealing factor to the combination result. 12, The apparatus of claim 11, whereia the sealing factor is ser bated ona least one of a mean arterial pressure (MAD) at the time of calibration, a diastolic blood pressure (DBP) fat the time of calibration, and a systolic blood pressure (SBP) atthe time of calibration 13. The apporats of claim IL, whersin the processor is farther configured to independently estimate the MAP. the DBP. and the SBPby adjusting atleast one ofthe weight, the Scaling factor, the fist change value, and the second change value 14, The apparatus of claim 1, wherein the processor is rer configured to estimate an amount of ebange in the blood pressure based on the relative change value of theUS 2020/0054290 AL feature and the offset, and estimate the blood pressure using the estimated amount of change and a reference blood pressure obtsined at the tint of calibration, 15. A method of estimate blood pressure, the method ‘comprising ‘measuring bio-sigal from a user: extracting a feature from the bio-signal st an extraction determining a change ina value ofthe feature that occurs during atime period between a caibeation time ofthe bio-signal and the extraction time; soquiring an offset based on the change in the value ofthe Teature; and estimating a blood pressure based on the change in the value of the feature and the ost. 16. The method of elaim 15, wherein the extracting the feature comprises extracting a cardiac output (CO) and total peripheral resistance (TPR) fom the biosignal, as the featue. 17. The method of claim 16, wherein the extracting the feature comprises acquiring, from the biosignal, character- istic points comprising atleast one of heat rate information, 8 shape of a wavelonm of the bio-signal, an area under the ‘waveform of the bio-signal, time and amplitude of a maxi- mum point of the bio-signal, time and amplitude of a tminimtim point ofthe bio-ignal and amplitade and time of ‘each of constituent pulse waveforms constituting the bio- Sgnal, and extracting the feature based on the acquired characteristic points 18. The method of claim 16, wherein the deteemining the change in the value of the feature comprise calculating = fist change valve by somaalizng the extracted CO based on ‘9 reference CO obtained at the time of calibration and ‘calculating socond change valve by normalizing the ‘extracted TPR based ona reference TPR obtained at the time of calibration, 19. The method of claim 18, wherein the determining the ‘change in the value of the feature frther comprises ealen lating a thied change value based on the fist change value snd the second change value and the acquiring the offset ‘comprises acquiring the offset by inputting the frst, the Second, and the third change values (0 an offset estimation model 20. The method of claim 18, wherein the feature further inchides feature related to pulse pressure 21. The method of claim 20, wierein the determining the change in the value of the feature futher comprises ealen lating a fourth change value by norliring the feature related based on pulse pressure to a feature related to pulse pressure obtained at the Hime of calibration and the acquiring the offset comprises inputting the fourth change value ito Feb. 20, 2020 an offset estimation model and acquiring te offset based on ‘an output result of the offset estimation model 22, The method of claim 21, wherein the determining the change in the valve of the feature furber comprises ealeu- Jating a third change value based on the first change vale ‘and the second change valve, and the acquiring the offset comprises acquiring the offaet by combining the thinl change value with the outpat result ofthe offset estimation ‘model, 23, The method of claim 18, wherein the estimating the blood pressure comprises applying a weight to the fist ‘change value and the second change valve, combining the ‘weighted frst and second change Values with the offset to obtain combination result, and applying a scaling factor to the combination res 724. The method of claim 18, wherein the estimating the blood pressire comprises estimating a smnount of cage in the blood pressure based on the change in the val of the feature and the offset, and estimating the blood pressure based on the amount of change and a reference blood pressure obtained at the tine of calibration, 28. An apparatus fr estimating blood pressure, the app: sats comprising 7 bio-sjgnal sensor configured to measure a bio-signal from a user ‘4 communication interface configured to receive a pulse ‘pressure of the user froma pulse pressure measurement apparatus; and 1 processor configured to extract a festure from the bio-signal, acquire an ake based on a relative change value of the received pulse pressure relative toa reference value ofthe pulse pressure atime of calibre tion, and estimate a blood pressure based on the relative change value of the feature and the offset 26, The apparatus of claim 25, wherein the processor is rer configured to extract, the feature, a eardise output (CO) and a total peripheral resistance (TPR) from the bio-signal, calculate relative change Values of the CO, the TPR. and the pulse pressure relative tothe ine of calibra tion by normalizing the CO, the TPR, and the pulse pressure, respectively, based on a reference CO, a reference TPR, aa ‘reference pulse pressure which are obtained at the time of calibration, 27. The apparatus of claim 26, wherein the processor is configured apply a weight tothe change value of the CO and the change value of the TPR, combine the weighted changed values of the CO and the TPR with the offset to ‘obiain a combination result, and estimate the blood pressure by applying a scaling factor to the combination result,