Hybrid Correlation-Neural Network Synergy for Gait Signal Classification - Advances in Heuristic Signal Processing and Applications - page 261

• Step length or length of one step,

•

Stride length or distance between two steps,

• Stride interval (stance, swing, and double support interval),

• Reaction force or force exerted by a person on the floor while walking,

• Orthopaedic angles or orientation of limb segments,

• Electromyographic (EMG) activity of the involved musculature during walking,

• Minimum foot clearance (MFC) during walking, during the mid swing phase of

the gait cycle.
Under the scope of the present work, benchmark gait signals that are freely avail-
able from the physionet database [ 35 ] have been utilized. The database contains
real-life gait signals acquired from both healthy and pathological subjects having
neurological disorders due to Parkinson's disease (PD), Huntington's disease (HD),
and Amyotrophic Lateral Sclerosis (ALS). The database was contributed by Haus-
dorff et al. [ 16 , 17 ], which includes 16 healthy subjects (2 men and 14 women)
aged 20-74 years, 15 PD subjects (10 men and 5 women) aged 44-80 years, 19 HD
subjects (6 men and 13 women) aged 29-71 years, and 13 ALS subjects (10 men
and 3 women) aged 36-70 years. Height and weight of the pathological subjects
recorded in the database were not significantly different from those of the healthy
subjects. This database also maintains a measure of disease severity or duration,
to indicate the extent to which a subject of the database is affected by PD, HD, or
ALS. The database uses Hohn and Yahr score for the subjects suffering from PD.
A higher value of this score indicates a more advanced condition of the disease.
The score varies from 1.5 to 4, for the PD subjects under consideration in this chap-
ter. For 60 % of these patients, the score is 3 or more, signifying a more advanced
state of the disease. The database uses total functional capacity measure for HD sub-
jects. Here, a lower score indicates more advanced functional impairment. The score
varies from 1 to 12 for HD subjects under consideration. Here, for almost 50 % of
these patients, the score is 5 or less (signifying more severe state of the disease), and,
for the rest 50 % patients, the score is more than 5. For the subjects suffering from
ALS, the severity measure maintained by the database is the time since the onset of
the disease. Here, for almost 80 % of the ALS patients, the severity of the disease is
moderate. The subjects were instructed to walk at their normal speed along a 77 m
long hallway. To measure gait rhythm and the timing of gait cycle, force-sensitive
resistors were placed as insoles in each subject's shoe. The gait time sequences were
obtained using these resistive sensors with output approximately proportional to the
force under the foot. Stride-to-stride measures of footfall contact times were derived
from these signals, and the stride time (i.e., the time from the initial contact of one
foot to the subsequent contact of the same foot) along with swing and stance times
was determined for each stride. For each subject, stride-to-stride measurements of
footfall contact times were acquired for approximately 300 s. In the present study,
the time sequences corresponding to the left and right stance intervals, the left and
right swing intervals, and the double support interval, each expressed as a percentage
of the stride time for each subject, are considered. Figures 12.1 , 12.2 , 12.3 demon-
strate the time sequence plots of the left swing interval, right stance interval, and
double support interval for some sample subjects. A close inspection of these plots