You are on page 1of 28

Sensor

based
monitoring
system
for
in-home Embedded Health Assessment for Senior
Citizens

By:
Sudhanshu Janwadkar, ME-II,VLSI & Embedded Systems
Guided By:
Dr. M. T. Kolte, PG Coordinator, Dept of EnTC, MITCOE

Content
Introduction
Earlier Work
Challenges in in-Home Health Assessment
System Block Diagram
Working: Feature Extraction

Principal Component Analysis

Fuzzy Pattern Tree

K-Nearest Neighbour

Neural Networks

Support Vector Machine

Working: Generating Health Alerts


Conclusion

Introduction

Recently, there has been an increased focus on technology


for enabling independent living and healthy aging.

Identification and assessment of health issues at early


stages, while they are still small, provides a window of
opportunity for curing the issues before they become
catastrophic

Older adults will benefit from early detection and


recognition of small changes in health conditions and get
help early when treatment is the most effective

Hence, we need an Unobtrusive, continuous monitoring in


the home for the purpose of assessing early health
changes

Introduction(contd..)

What is in-Home Embedded Health Assessment?

Sensors embedded in the environment capture


behaviour and activity patterns.

Changes in patterns are detected as potential signs


of changing health.

Based on the features extracted from in-home


sensor data, health alerts are generated to
clinicians

Clinicians analyze each alert and provide a rating


on the clinical relevance.

Introduction(Contd..)

These ratings are then used as ground truth for


training and testing classifiers.

Thus, this system is a health change detection


model

Thus, this approach provides a method for


detecting health problems very early, so that
early treatment is possible.

This method of passive in-home sensing also


alleviates compliance issues.

Earlier Work
Researcher

Research Conclusion

J. A. Kaye et
al. [1]

Both daytime and night time activity have been


investigated using in-home sensors.
Passive infrared (PIR) motion sensors have been used
to capture activity in a particular location in the home

M. Chan et al.

The pattern of room to room activity has been studied


as a means of investigating health changes

[2]

P. Cuddihy et
al., [3]

Motion density from PIR motion sensors (i.e., number


of events per unit time) capture overall activity level
and these have been linked with health conditions

Earlier Work(Contd..)
Researcher

Research Conclusion

T. van Kasteren sleep patterns have been studied using


et al. [4]
motion sensors bed mats or load cells
D C. Mack et
al. [5]

The detection of cognitive changes, using a


combination of motion, bed and door sensing,
medication tracking and a phone sensor for
detecting incoming and outgoing calls

M. MonteroOdasso et al.
[ 6]

Walking speed has been captured using


motion sensors ,video, radar and depth
images. Walking gait has been linked to both
physical and cognitive health

Challenges in in-Home Health Assessment

Identification of best parameters to track for health change;


some parameters may be too late for very early health
change detection.

Many seniors have multiple chronic health conditions to manage


and the interaction may present changes in a variety of ways.

Many seniors refuse to wear them, forget to charge batteries, or


are unable to operate them

Many wearable sensors are still at the prototype stage and not
yet ready for long-term use in an unstructured home
environment.

Challenges still need to be addressed, such as unobtrusiveness,


miniaturization, and robustness before they can be adopted

System Block Diagram

System Block Diagram (Explanation)

An Embedded sensor network collects data on behaviour


and activity patterns.
For example:

Passive infrared (PIR) motion sensors are used to capture motion


in a room area and also for localized activity, e.g., in the refrigerator,
in kitchen cabinets, on the ceiling over the shower etc

Bed Sensors are installed beneath the bed and on the bed to capture
events such as low or high pulse rate, low or high respiration rate etc

Sensors for detecting every minute change in activity and behaviour


pattern are installed everywhere in home

Data from sensors installed in apartments are logged and


stored on a secure server

A one-dimensional (1-D) alert algorithm is used to


generate health alerts.

System Block Diagram (Explanation)

When is a health alert generated?


Example:

The PIR motion sensors generate an event every seven


seconds if there is continuous motion.
This is used as an artifact to capture the general activity level in
the home by computing a motion density as motion events per
unit time.
For example, a resident with a sedentary lifestyle may generate
only 50 motion events per hour, whereas a resident with a very
active life style may generate 400 or more motion events per
hour

Clinicians analyze each alert using an electronic health


record (EHR) based on the experience and patient history.

Based on their clinical expertise, they rate the clinical


relevance of the alert.

Complexity of Captured Data

The fundamental observation of such in-home


Embedded Health Assessment System :
Huge complex web of sensors
A sensor to capture change in each activity
and behaviour pattern
Further, each activity may be classified as
High, Medium or Low OR
Increasing or Decreasing
The captured activity pattern is a Ndimensional feature space

Complexity of Captured Data

For example, consider only following four alert


parameters: bathroom activity, bed restlessness,
kitchen activity, and living room activity; are
considered.

If both increasing and decreasing changes (the


current day's count compared to the baseline
period) are considered for all three time periods
(daytime, night time, and full day), the
dimensionality of the feature space is 24.

Consider what if all sensors are taken into account

Analyzing the Feature Space

In analyzing the health alerts, it is observed that some


of the parameters do not typically contribute to alerts

Others generate a few alert but their information is not


enough to be used for supervised learning.

Considering both increase and decrease in activity


pattern can be redundant in some cases.

Thus, observation and experience can be sufficient in


some cases to reduce feature space.

For others, we require complex methods like Principal


Component Analysis etc

Principal Component Analysis

Principal component analysis (PCA) is a statistical


procedure that uses an orthogonal transformation
to convert a set of observations of possibly
correlated variables into a set of values of linearly
uncorrelated
variables
called
principal
components.
The number of principal components is less than
the number of original variables.

Analyzing Feature Space

K Nearest Neighbour
Fuzzy Pattern Tree
Neural Network
Support Vector Machine

K-Nearest Neighbour

In pattern recognition, the k-Nearest Neighbours algorithm


is a method used for classification and regression.

The input consists of the k closest training examples in the


feature space. The training examples are vectors in a
multidimensional feature space, each with a class label.

In k-NN classification, the output is a class membership.

An object is classified by a majority vote of its neighbours,


with the object being assigned to the class most common
among its k nearest neighbours

kis a user-defined constant

K-Nearest Neighbour (contd..)

Example:

K-Nearest Neighbour (contd..)

Choose Odd Value of K for 2 value problem


K must not be a multiple of number of classes
The main disadvantage of K-NN method is
difficulty in searching the nearest neighbours
for each sample.

Application
Each activity pattern is considered as a input
training pattern
Health Alert is a output class

Fuzzy pattern tree (FPT)

Fuzzy logic is a form of many-valued logic in which the truth


values of variables may be any real number between 0 and
1, considered to be "fuzzy".

Fuzzy logic has been employed to handle the concept of


partial truth, where the truth value may range between
completely true and completely false.

Fuzzy logic involves linguistic variables(do not take


numerical values)

Degree of output may


(membership) functions.

Fuzzy Rules are framed to determine the class of output.

be

managed

by

specific

Fuzzy pattern tree (Contd..)

Example:

The meanings of the expressions cold, warm, and hot are


represented by functions mapping a temperature scale.

A point on that scale has three "truth values" one for


each of the three functions

Since the red arrow points to zero, this temperature may be


interpreted as "not hot". The orange arrow (pointing at 0.2)
may describe it as "slightly warm" and the blue arrow
(pointing at 0.8) "fairly cold".

Fuzzy pattern tree (Contd..)


A

fuzzy pattern tree is a method that uses


domain knowledge and does not require
training.

Application
The output is as follows:
IF Bathroom activity for the full day is an Increase
OR Bathroom activity at night time is an Increase
OR Bed restlessness for the full day is an Increase
OR Bed restlessness at night time is an Increase
OR Kitchen activity at night time is an Increase
OR Living room activity at night time is an Increase
THEN Alert is Clinically Relevant

Gaussian-based

membership functions are


used for the input parameters.

Neural Network
In supervised learning, we are given a set of example pairs
{ (x , y); x X, y Y} and the aim is to find a function { f:X
Y} in the allowed class of functions that matches the
examples.

In other words, we wish to infer the mapping implied by the


data; the cost function is related to the mismatch between
our mapping and the data and it implicitly contains prior
knowledge about the problem domain.

A commonly used cost is the mean-squared error

When one tries to minimize this cost using gradient descent


for the class of neural networks using well-known backpropagation algorithm

Support Vector Machine

Support vector machines are supervised learning models with


associated learning algorithms that analyze data used for
classification and regression analysis.

Given a set of training examples,, an SVM training algorithm


builds a model that assigns new examples into one category or
the other

A non-probabilistic binary linear classifier.

A support vector machine constructs a hyper-plane or set of


hyper-planes, which can be used for classification and
regression

Using Kernel trick, possible to classify inputs which are not


linearly separable in that space.

Generating Health Alerts

The logged sensor data are automatically analyzed on a daily


basis, looking for changes in an individual's data patterns.

If a change is detected for the current day, an alert email is


sent to clinicians.

Clinician determine whether the alert is relevant for this


resident from a clinical perspective.

Clinician rates the clinical relevance of the alert on a five


point scale, from 1 (not clinically relevant) to 5 (very clinically
relevant).

This data is used for training and learning of the methods


discussed and further development of the alert algorithms.

Conclusion

The importance of sensor data for capturing early signs of


health decline.

Unobtrusive Continuous Time In-Home Embedded Health


Assessment System

Identifying health decline early provides a window of


opportunity for early treatment and intervention that can
address health problems before they become catastrophic.

This offers the potential for improved health outcomes,


reduced healthcare costs, continued independence and
better quality of life.

Thank-you

References

Marjorie Skubic, Rainer Dane Guevara, Marilyn Rantz, "Automated Health Alerts Using In-Home
Sensor Data for Embedded Health Assessment", IEEE Journal of Translational Engineering and
Medicine, Volume 3, 2015,pp2168-2372

1.

J. A. Kaye et al., ``Intelligent systems for assessing aging changes: Homebased, unobtrusive,
and continuous assessment of aging,'' J. Gerontol., Psychol. Sci., vol. 66B, no. 1, pp. i180i190,
2011.

2.

M. Chan, E. Campo, and D. Esteve, ``Assessment of activity of elderly people using a home
monitoring system,'' Int. J. Rehabil. Res., vol. 28, no. 1, pp. 6976, 2005.

3.

P. Cuddihy et al., ``Successful aging,'' IEEE Pervasive Comput., vol. 3, no. 2, pp. 4850, Apr. 2004.

4.

T. van Kasteren, A. K. Noulas, G. Englebienne, and B. Krse, ``Accurate activity recognition in a


home setting,'' in Proc. 10th Int. Conf. Ubiquitous Comput., 2008, pp. 19.

5.

D. C. Mack, J. T. Patrie, P. M. Suratt, R. A. Felder, and M. Alwan, ``Development and preliminary


validation of heart rate and breathing rate detection using a passive, ballistocardiography-based
sleep monitoring system,'' IEEE Trans. Inf. Technol. Biomed., vol. 13, no. 1, pp. 111120, Jan.
2009.

6.

J. Kaye et al., ``Unobtrusive measurement of daily computer use to detect mild cognitive
impairment,'' Alzheimer's Dementia, vol. 10, no. 1, pp. 1017, 2014.

7.

M. J. Rantz, ``Evaluation of health alerts from an early illness warnin system in independent living,''
Comput., Inform., Nursing, vol. 31, no. 6, pp. 274280, 2013.

You might also like