基于Haar小波和融合HMM的步态识别方法

摘要

This paper presents a novel gait recognition approach based on Haar wavelet and fused hidden Markov model. It solves the problem that in gait recognition there are insufficient key points of the gait feature in each region. First, the approach converts images from video sequences to binary contour, and uses Haar wavelet transform to obtain the distinct key points of gait features. Then two sub-images are utilised to represent the gait feature of each contour, and the principal component analysis is employed to reduce the number of dimensions. Finally, fused hidden Markov model is used for training and testing. Simulation result indicates that the approach can simplify the process of gait identification, and can also improve the recognition accuracy.%为解决步态识别中每个区域的步态特征要点匮乏问题,提出一种基于Haar小波及融合的隐马尔可夫模型Fused-HMMs(fused hidden Markov models)的步态识别方法.该方法首先把视频序列中的图像转换成二进制轮廓,利用Haar小波变换取得显著的步态特征要点；其次采用两个子图像来表示各个轮廓的步态特征,并通过主成分分析法减少维数；最后,利用融合HMM进行训练和测试.仿真结果表明该方法不仅可以简化步态辨识过程,而且还能够提高识别准确率.