Gait Recognition Based On Width And Angle Features

In modern society, there are more requirements on the accuracy, security and practicability of human recognition. The traditional methods are not very adapted. Biometrics is playing an important role now. Gait recognition is one of these techniques.Human gait recognition is the process of identifying individuals by their walking manners. It consists of three parts: preprocessing of gait sequences, feature extraction and classification. Feature extraction is the most important thing on which our study focuses.In this thesis, we construct the background through three frames differencing, subtract it from the current image to get the foreground. Then we extract the width features from human body, use Radon transform to analyze the low limps and get the limp angle features. After the PCA transform, we fuse the features to train and recognition. One method bases on DTW, and the other uses HMM classifier to get the model of every class in the database. Finally, to recognize the unknown sample, we calculate the output probability of it to all classes and choose the person corresponding to the maximum output probability as its classification.The methods can capture the structural and the dynamic characteristics that are unique to one individual, decrease the affection of self-occlusion and shadow. The way to extract features is robust to the change of speed, also simpler and faster than methods based on models. The experimental results on two general databases demonstrate that the methods are effective.