Digital Signal Processing Reference
In-Depth Information
12.1 Introduction
Going by the lexicon, the term 'gait' represents the manner of moving on foot.
Movement is an important routine activity of all human beings. Any impediment
in movement, substantially downgrades the quality of our life. On the other hand,
human gait signal has the potential to serve as an important biometric trait pri-
marily due to its inconspicuousness, as it can be acquired from a distance with-
out the prior knowledge of the subject [
5
]. The investigations related to gait as
a distinguishing feature were first attempted a few decades back, from a medi-
cal/behavioral viewpoint [
10
,
20
]. This was followed by several attempts to investi-
gate the problem of gait recognition, in the context of capturing and analyzing gait
signals [
9
,
19
,
27
,
30
,
33
].
So far as medical applications are concerned, the domain of gait and human
movement science has stolen much of the limelight with the appreciation of the
fact that locomotor dysfunctions demand considerable medical attention, involve
high costs of treatment, and may also turn out to be fatal in certain cases [
25
].
At a certain point of time, statistics revealed that 90 % of the adults with cerebral
palsy (CP) in the U.S.A. lacked access to periodic health checks-up [
31
], although
more than 50 % of the CP hemiplegics were required to have constant personal
assistance [
43
]. The elderly people face progressive gait disorder, which enhances
the possibility of death due to falls and bone fractures [
46
]. Half of these victims
of fall, if left unattended for more than two hours, are exposed to the danger of
succumbing to dehydration, hypothermia, pneumonia, pulmonary embolism (which
accounts for 38 % of deaths in hip fracture falls), rhabdomyolysis (which is a toxic
breakdown of muscle fibres), and pressure ulcers [
2
].
A large number of research workers have examined gait signals by different
methods. The methods put forward so far are primarily aimed at clustering gait
signals into young and old categories [
1
,
32
,
34
,
47
]. Reported works on analysis of
gait signals for identifying neurological disorders in subjects, and also distinguish-
ing them from healthy subjects, have been few and far between. The use of artificial
neural networks (ANNs) for automated identification of gait patterns has already
been reported [
1
,
3
]. The work referred to considered eight subjects under three gait
conditions, namely, normal gait, a simulation of leg length difference, and a simu-
lation of leg-length difference [
1
]. The features from hip-knee joint angle diagrams
were utilized to train the ANN, which was subsequently used for identifying the
type of gait. A three-class classification problem solution yielded a classification
ratio of 83.3 %.
Young-elderly classification of gait signals plays a significant role in identifying
the onset of gait related disorder in aged people, so that preventive measures can be
taken against fall [
14
,
32
]. In an investigation [
3
], statistical features were processed
by support vector machine based classifier for binary classification of gait signals.
Twenty-four such features were derived from the minimum foot clearance (MFC)
data of 58 subjects to obtain classification into young and elderly gaits. A mean
classification accuracy of 83.3 % was achieved.
Automated determination of whether a subject is suffering from neurological dis-
eases, and also the type of disease, is another object of interest in investigations on
Search WWH ::
Custom Search