Research On Gait Recognition Based On The Features Of Static Plantar Pressure

In recent years, along with the rapid development of biological recognition technology, gait feature which is a new kind of biological feature has gradually becoming a frontier in the research of pattern recognition. Traditional methods of gait recognition are based on the video stream data, through the camera to collect people walking posture. The research of this gait recognition is long, and result quite abundant, but it is easily affected by background environment and the shelter of clothes. In recent years, a new kind of gait recognition based on plantar pressure data, which overcomes the shortcomings of traditional methods, can be used in many occasions.This paper combines with the theory of mechanical, biology and pattern recognition, and the relevance algorithms are applied to the classification of gait data based on the plantar pressure, constructing the relevance feature space, and the recognition effect of plantar pressure feature is tested through the experiments of real data. The main work and research results are as follows:1. Analyzes the current research of gait recognition. Gait recognition is mainly including two kinds:one is based on video data streams, and the other plantar pressure data. This paper introduces the main content of the two kinds of gait recognition in turn, and the research methods and important applications in the major domestic and foreign literature.2. Introduces the collection process of the plantar pressure data. Firstly, this paper introduces several common collection equipments, including their advantages and disadvantages; secondly, introduces the collection equipment in this paper and the data processing software in detail; lastly, uses the equipment to collect 20 individual’s static plantar pressure data for research.3. Proposes a method of gait recognition based on Hu invariant moments and support vector machine (Support Vector Machine, SVM). The method extracts the Hu moment features of plantar pressure image at first, constructing a 7-dimensional matrix of the plantar pressure features to describe the contour and detail information of the image; secondly, normalizes the Hu moments’ feature for the different of the magnitudes between the 7-dimensional features in Hu moments; finally, test the classification performance at the database which had been set up by S VM.4. Proposes a gait clustering algorithm based on low dimensional plantar pressure features. For the current methods mostly based on the foot’s plantar pressure and shape features, this paper synthesizes the classic features to construct a feature vector; then eliminates the redundant information between these features using non negative matrix method (Non-negative Matrix Factorization, NMF); combines with the fuzzy c-means algorithm (Fuzzy c-means Algorithm, FCM) to classify the low dimensional feature; finally compares with the features in the traditional methods to test the effectiveness of the algorithm.