Abstract:

BACKGROUND: The gait data is an objective parameter to patients who have musculoskeletal disorders or nervous system diseases. It can evaluate recovery of illness and set up treatment method. How to classify the gait data effectively has become an important research topic.
OBJECTIVE: In order to diagnosis illness effectively and provide scientific basis for setting up treatment method, using the modified support vector machines algorithm to classify gait data.
METHODS: Modified Support Vector Machines algorithm was proposed, and 720 samples were selected from 6 group data aged 40-59 years from Gait Dynamics in Neuro-degenerative Disease Data Base. Gait data models were established using left swing interval and left stance interval. After normalization, data were preprocessed with Fuzzy C-Mean algorithm, and then classified gait data utilizing support vector machines. The classification ability was verified by support vector machines algorithm with various kernel functions.
RESULTS AND CONCLUSION: By comparing classifiers using different kernel function, the result is that the classifier with modified support vector machines algorithm can classify small sample size gait data and set up treatment method effectively.