Computational Intelligence in Gait and Movement Pattern Analysis - Computational Intelligence in Biomedical Engineering - page 309

Diplegia. This is a type of CP that affects both sides.

Hemiplegia. This is the mildest form of CP that affects only one side
of the body. An individual with hemiplegic type of CP will have only
a few limitations while performing his/her daily activities.

Increased lumbar lordosis. Lordosis is used to aid walking.

Lateral trunk bending. Painful/abnormal hip joint, wide walking base,
and unequal leg length.

Parkinsonism. Degeneration of the basal ganglia of the brain that
often results in shuing gait.

Posterior trunk bending. Compensation for weak hip extensors early
in the stance phase.

Quadriplegia. Involvement of the four limbs, the trunk, and the head.

7.4 CI Techniques in Gait Analysis
7.4.1 Artificial Neural Networks
CI techniques including NNs, fuzzy logic, genetic algorithms, and SVMs have
been applied to model gait data and to classify gait types from their features
(Chau, 2001; Schollhorn, 2004; Begg et al., 2005b). Of these techniques, NN
applications are the most widespread and most of these applications focus on
gait pattern classification and recognition. There have also been attempts to
use CI techniques for gait data modelling, for example, Sepulveda et al. (1993)
successfully applied NNs to map lower-limb joint angles and joint moments
onto electromyographic activity patterns.
A number of investigations have used NNs for the automatic identifica-
tion of gait types, such as normal, abnormal, or pathological, using features
from gait recordings. Figure 7.11 illustrates steps in a gait classification task
based on NNs. Such an automated model could lead to useful applications,
for example, in the diagnosis of gait abnormalities and assessment of gait
following treatments and interventions.

Neural
network
classifier

Gait
class

Gait data
collection

Gait
feature
extraction

FIGURE 7.11
Main steps in a gait classification task using neural networks.