Real-time gait classification based on fuzzy associative memory

Gait classification has a potential to be used for recognition. This paper describes a method for classifying the gaits of human bodies in video sequence and deals with the classification of human gait types based on the notion that gait types can be analysed into a series of consecutive posture types. First, according to the different sorts of movements, we make a set of standard image contours using recursion method and put them into the database. Through the hidden Markov models (HMM), different behaviour matrices based on spatio-temporal are acquired. Then, according to the video sequence, silhouettes are extracted using the background subtraction. A moment distance method is presented to obtain the similarity degree of silhouettes, which is estimated by comparing the incoming silhouettes to the database silhouettes. Finally, fuzzy associative memory (FAM) classifier is proposed to infer the gait classification of a walker. An evaluation of ten kinds of gaits involving walk, stand, faint, sit, run, bench, jump, crouch, wander and punch are given. The experiment tests show some encouraging results, which indicate that the method can be a choice for solving the problem described although more tests are required.