Professional Documents
Culture Documents
A R T I C L E I N F O A B S T R A C T
Keywords: With the widespread of Internet of things, varied innovative tools are required to control the nearby smart
BCI communications devices. In this contest, brain-computer interface (BCI) technology can empower individuals to directly control
IoT varied electronic devices such as those used in smart home/office and associative robots via their thoughts. This
I/Q down converter
process requires efficient transmission of electro-corticographical (ECoG) signals from implanted electrodes in
LMS algorithm
Time-interleaved ADCs
side the brain to an external receiver located outside on the scalp. However, the realization of this vision is still
challenging task due to the limited reliability of adopted wireless BCI communication systems. In particular, the
generated artifacts due to in-phase/quadrature (I/Q) imbalance of utilized down converter and time interleaved
analog-to-digital converters (ADCs) may lead to significant interference to desired signal which affects the
detection performance. In this paper, an efficient low complexity balance technique is proposed for BCI com
munications to mitigate the interference using the adaptive least mean square (LMS) algorithm. Performance
results of conducted experiments on a near-realistic phantom human brain model are shown to validate the
designed balanced BCI (BBCI) scheme compared with the existing BCI approach.
1. Introduction machines, and other modern smart home applications. This will make
the home/office environment more accessible and comfortable for
During the last decade, intensive research efforts have been per disabled persons as well as normal users by simply employing their
formed on the promising brain-computer interface (BCI) technology for neuronal signals (i.e. brain reactions and thoughts) [7–10].
various health applications. It is used mainly to facilitate direct
communication links between the central nervous system of human 1.1. Technical background
beings with the external electronic devices. This approach can be
considered as a novel extension to the initial objective of BCI, repre Most of persons with tetraplegia have flawless cerebrum work but
sented by the realization of immediate communication environment cannot move due to some damage/infection that influences the spinal
between the brain and embedded devices, to help patients suffering from cord. Therefore, different BCI schemes are developed to sidestep the
spinal cord injuries, paralysis situations, and amyotrophic lateral scle harmed nerves and build up an alternative connection between brain
rosis [1–5]. In practice, BCI schemes have the effects to reduce the need signals and embedded devices to re-establish muscles’ development or
for typical information delivery methods and then improve the lifestyle co-appointment of paralyzed muscles [7,11]. For instance, Fig. 1(a)
of disabled persons. shows how the bypass wireless channel can handle the wireless trans
On the other hand, massive number of devices/machines can be mission of brain messages as encoded data through BCI to activate and
connected to the Internet, usually referred as the Internet of things (IoT), control the movement of paralyzed arm through external controller
allowing a fundament shift on the way that individuals can interact with (brain spine interface, BSI) or implanted device. The implementation of
the surrounding objects [6]. Therefore, the integration of IoT and BCI utilized BCI requires bio-compatible electrodes for signal interface with
schemes represents an emerging technology to control varied personal neurons in the brain’s cortex and an external transceiver device on the
phones and computers, electronic and wearable devices, smart scalp to receive the electro-corticographical (ECoG) signals and
* Corresponding author.
E-mail addresses: salajraw@eng.ucsd.edu (S. Al Ajrawi), alhussaibi@stu.edu.iq (W. Al-Hussaibi), rrao@ucsd.edu (R. Rao), msarkar2@mail.sdsu.edu (M. Sarkar).
https://doi.org/10.1016/j.imu.2019.100276
Received 27 October 2019; Received in revised form 26 November 2019; Accepted 29 November 2019
Available online 16 December 2019
2352-9148/© 2019 Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
S. Al Ajrawi et al. Informatics in Medicine Unlocked 18 (2020) 100276
retransmit them to intended controller on the defective part of disabled found when extracting one or more channels in multi-channel filter
person’s body [3]. banks. This will have direct impact to improve the performance of BCI
On the other side, BCI can be considered as an emerging alternative system significantly.
for supporting interaction between IoT objects and individuals as shown
in Fig. 1 (b). In this scenario, the ECoG signals of wheelchair user are 1.2. Aims and contributions
transmitted with high data rate to the implanted external transceiver on
the scalp and then to the analyzer device (such as computer, tablet, and The employment of signal processing techniques to undo the signal
smart phone) for data analysis and examinations with pre-portrayed degradation caused by the analog I/Q imbalance is adopted as an effi
examples. The determined person’s order is then sent through IoT cient solution in varied electrical, electronic, and communication sys
network to the intended smart devices/applications for desired actions. tems [15–20]. This comes against the common idea that this kind of
In any of the aforesaid BCI applications, information messages are distortion is challenging and irreversible [14]. In this paper, we aim to
transmitted with high data rate using one of the well-known signal mitigate the inherited I/Q imbalance in BCI communication systems
modulation methods such as quadrature phase shift keying (QPSK) to using low complexity balance technique to achieve the target perfor
the receiver side [11–13]. Besides, efficient BCI communication system mance and exploit this important technology.
can be determined typically when a distortion free transmission from the The main contributions of this work are summarized as follows:
wireless transmitter inside the brain to the receiver end on the scalp is
achieved. However, realization of this target is difficult owing to the � An efficient low complexity balance technique is proposed for BCI
effect of in-phase/quadrature (I/Q) imbalance in the employed fre communications to mitigate the interference caused by I/Q imbal
quency down converter and time interleaved analog-to-digital con ance using the adaptive least mean square (LMS) algorithm. The
verters (TI-ADCs) at the receiver processor. In particular, the gain and integration of these cost-effective approaches is very important for
phase imbalance in the two paths of analog processor can cause unde feasible BCI applications toward wide adoption of this technology.
sired coupling between the positive and negative frequency components � For the considered technique, we present the time model of I/Q
of the signal carried by these paths [14–22]. This may potentially lead to mismatch in BCI and the employed adaptive cancellers to correct the
considerable interference level to desired signal which affects the DC offsets and imbalance defects of frequency down converter and
detection performance, and hence the required response. Therefore, QPSK signalling. A balanced model is demonstrated then to cancel
controlling this kind of mismatch is very important in the utilized re the DC offsets and spectral artifacts produced by the pair of TI-ADCs.
ceivers that detect signals spanning a wide range of levels as might be
Fig. 1. BCI applications: (a) Spinal cord using by-pass wireless channel; (b) Brain control for smart devices and IoT network.
2
S. Al Ajrawi et al. Informatics in Medicine Unlocked 18 (2020) 100276
� Performance outcomes of conducted tests on a near-realistic phan reader is supported with an antenna and Ethernet connection for data
tom model of human brain demonstrate the effectiveness of designed transmission to intended controllers such as laptops/smartphones.
balanced BCI (BBCI) scheme compared with the generic BCI in Refs.
[3,11]. 3. Proposed balance technique for BCI
The rest of this paper is organized as follows: In Section 2, the system For QPSK signalling, a balanced modulator is required which can be
model of BCI communications is described. Section 3 presents the pro comprised of two matched amplitude modulation modules. However,
posed balance technique for BCI applications. Section 4 deals with maintaining perfect balance between these modulation modules over RF
conducted experiments and performance results. Discussion on the carrier is not an easy process in BCI communication system. This kind of
achieved results is presented in Section 5. Finally, Section 6 concludes imbalance between I/Q channels may result in a signal distortion
the paper. leading to undesired inter channel interference (ICI) and increased error
rate performance [12]. For instance when a target bit error rate of 10 6
2. System model of BCI communications is desired, the required levels of amplitude imbalance and phase
imbalance should not exceed 1 dB and 5o , respectively [15,16]. There
In this work, we consider a wireless BCI communication system for fore, an efficient low complexity balance method is proposed in the
health applications as shown in Fig. 2. It consists of implanted BCI following.
transmitter unit, wireless channel, and receiver device. The brain ECoG
signals from implanted electrodes are modulated using QPSK technique,
and the generated radio frequency (RF) signals are transmitted then over 3.1. IQ down converter
wireless channel through brain tissues to the fixed receiver on the scalp.
The received RF signal with an additive white Gaussian noise compo The received QPSK ECoG signal in BCI is processed using IQ down
nent is pre-filtered as in the typical communications systems to elimi converter which consists of two branches of multipliers, low pass filters
nate the noise effects [23]. Then, it will be down converted and sampled (LPFs), and ADCs as shown in Fig. 3(a). The in-phase and quadrature
using ADC unit for signal demodulation. The demodulated brain signals components for down conversion are represented by I ¼ cosðωtÞ and Q ¼
are processed to prepare the information and control messages for sinðωtÞ, respectively, where ω ¼ 2πf, and f denotes the carrier fre
response/analyzer device, while a parameter estimation and balance quency. The mismatch in IQ down converter which leads to ICI is shown
unit is used to solve the problem of I/Q imbalance in BCI system. in the second path due to phase distortion ðαÞ and gain distortion ð1 þεÞ
Note that the propagation delay due to wireless signal propagation elements. The time-domain model of I/Q mismatch is demonstrated in
between the implanted electrodes and the reader on the scalp through Fig. 3(b). This model shows also the insertion of DC components in both
human tissues can be mitigated using multiple in-brain transmitters and of in-phase and quadrature paths due to the utilized ADCs. This can be
relevant multiple access technique as in Ref. [3]. This has direct impact explained mathematically without showing the time index n as
to satisfy the required user experience. But, signal propagation delay
between mounted device on the scalp and computer/smart devices I ’ ¼ I þ DCI (1)
through indoor wireless channel has no significant impact on the per
formance of BCI as in the existing wireless systems [24]. Moreover, other Qj ¼ Qð1 þ 2 Þ þ αI þ DCQ (2)
signal processing delays can be ignored based on the advancement in
where α ¼ η sinðφÞ; ε ¼ η cosðφÞ 1; η and φ represent the amplitude
digital signal processing chips [4].
and phase elements of distortion signal, respectively. So, the quadrature
As in our previous work [3], a wireless identification sensing plat
component can be simplified and rewritten as
form 5 (WISP5) is used for the adopted BCI system. It is based on a
passive RF identification (RFID) tag that works on an ultra-low power Q’ ¼ η sinðωt þ φÞ þ DCQ (3)
and micro-controller for sensing, processing, and communication [25].
The receiver on the scalp is implemented using RFID reader such as It can be seen that the I and Q signal components are not orthogonal due
Impinj’s Speedway model [26]. This RFID model is used to detect the to gain and phase imbalance as well as DC insertion in both of signal
WISP signals and also emits RF signals (in the frequency range 908–928 paths. Therefore, α and ε elements should be estimated, as b α and bε , for
MHz) to send commands and power to the implanted WISP tags. The possible signal correction with DC cancellation.
3
S. Al Ajrawi et al. Informatics in Medicine Unlocked 18 (2020) 100276
n� o n� o
� �
C ε ðnÞ ¼ E �eε ðnÞj2 ¼ E �bε ðnÞ εðnÞj2 (5)
where Ef:g is the expectation (mean) operator while eα ðnÞ and eε ðnÞ
denote the errors at the current time sample n. The parameters α b ðnÞ and
ε ðnÞ are used then to find bIðnÞ and QðnÞ
b b signals, respectively.
4
S. Al Ajrawi et al. Informatics in Medicine Unlocked 18 (2020) 100276
Fig. 6. Learning curves of the in-phase and quadrature DC components (amplitude in volts) of BBCI as a function of time samples.
Fig. 7. Parameter estimation based on LMS algorithm: (a) Phase estimate in radian as a function of samples; (b) Gain estimate in volts as a function of samples.
The impact of ICI in BBCI scheme with QPSK signalling is shown in DC cancellation as a function of the number of received time instance
Fig. 8 through Lissajous patterns and constellation diagrams of detected for: (a) first 2000 samples and (b) last 2000 samples. The spectrum of
signals. From Fig. 8(a) of Lissajous patterns, it can be seen clearly that output signal of balanced model is demonstrated also in (c) for the first
both of phase and gain components (distorted BCI signals) are corrected 2000 samples and (d) for the last 2000 samples. These figures demon
by the end of balancing process using BBCI scheme. This is shown also in strate that the first samples will suffer from possible signal distortion due
Fig. 8(b) where a clear QPSK constellation is achieved for BBCI system to imperfect estimation of the system parameters. But, the system will
after cancellation of the DC offset, phase error, and gain imbalance. converge to best estimate and produce reliable performance as the time
The spectrum (in log magnitude) of TI-ADCs is shown in Fig. 9 after samples grow.
5
S. Al Ajrawi et al. Informatics in Medicine Unlocked 18 (2020) 100276
Fig. 8. QPSK of distorted (BCI signals) and non-distorted signal components (BBCI signals): (a) Lisasajous patterns; (b) Signal constellation.
6
S. Al Ajrawi et al. Informatics in Medicine Unlocked 18 (2020) 100276
Fig. 9. Spectrum (log magnitude) of TI-ADCs after dc cancelling as a function of time instance for: (a) first 2000 samples and (b) last 2000 samples. Spectrum of
output signal of balanced model for (c) first 2000 samples and (d) last 2000 samples.
7
S. Al Ajrawi et al. Informatics in Medicine Unlocked 18 (2020) 100276
[10] Teles A, Cagy M, Silva F, Endler M, Bastos V, Teixeira S. Using brain-computer [19] Johansson H. A polynomial-based time-varying filter structure for the
interface and internet of things to improve healthcare for wheelchair users. In: The compensation of frequency-response mismatch errors in time-interleaved ADCs.
eleventh international conference on mobile ubiquitous computing, systems, IEEE J Sel Topics Signal Process Jun. 2009;3(3):384–96.
services and technologies; 2017. [20] Tsui KM, Chan SC. New iterative framework for frequency response mismatch
[11] Al Ajrawi S, Mohanty S, Sarkar M, Rao R, Mihovska A, Baweja H. Investigating correction in time-interleaved ADCs: design and performance analysis. IEEE Trans
feasibility of multiple UHF passive RFID transmitters using backscatter Instrum Meas Dec. 2011;60(12):3792–805.
modulation. Bellevue, WA: SSMC; Jul 2017. p. 552–63. [21] Soudan M, Vogel C. Correction structures for linear weakly time varying systems.
[12] Harris F, Parekh S, Gurantz I. I-Q balancing techniques for broad-band receivers. IEEE Trans Circuits Syst I Sep. 2012;59(9):2075–84.
In: Proceeding of the Software Defined Radio Technical Conference; 2005. 6 pages. [22] Tsui KM, Chan SC. A novel iterative structure for online calibration of M-channel
[13] Al-Hussaibi W. Signal design for higher capacity correlated fading signature time-interleaved ADCs. IEEE Trans. Instrum. Meas Feb. 2014;63(2):312–25.
multiple access. Wirel Netw Nov. 2014;20(8):2275–85. [23] Al-Hussaibi W. Effect of filtering on the synchronization and performance of chaos-
[14] F. Harris, “Artifact-corrected time-interleaved ADC,” U.S. Patents, US8957798B1, based secure communication over Rayleigh fading channel. Commun Nonlinear Sci
17 pages, Feb. 2015. Numer Simul Sep. 2015;26(1–3):87–97.
[15] Tsai T, Hurst P, Lewis S. Time-interleaved analog-to-digital converters for digital [24] Al-Hussaibi W, Ali F. Generation of correlated Rayleigh fading channels for
communications. Proc. IASTED Conf. Circuits, Signals Syst 2004:193. accurate simulation of promising wireless communication systems. Simul Model
[16] Liu C-L. Impacts of i/q imbalance on qpsk-ofdm-qam detection. EEE Trans Consum Pract Theory Jun. 2012;25(4):56–72.
Electron Aug. 1998;44(3):984–9. [25] Sample A, Yeager D, Powledge P, Mamishev A, Smith J. Design of an RFID-based
[17] Tsou H. The effect of phase and amplitude imbalance on the performance of offset battery-free programmable sensing platform. IEEE Trans. Instrum. Meas Nov.
quadrature phase-shift-keyed (OQPSK) communication systems. The 2008;57(11):2608–15.
Telecommunications and Mission Operations Progress Report; Sep. 1998. p. 1–15. [26] IMPINJ. Speedway R420 RAIN RFID Reader. URL: https://www.impinj.
[18] Vogel C, Mendel S. A flexible and scalable structure to compensate frequency com/platform/connectivity/speedway-r420/.
response mismatches in time-interleaved ADCs. IEEE Trans Circuits Syst I Nov. [27] Monson H. Hayes, statistical digital signal processing and modeling. first ed. Wiley;
2009;56(11):2463–75. 1996.