基于小波变换和支持向量机的步态识别算法

Gait Recognition Based on DWT and SVM

为了快速准确地进行人体运动步态识别,基于运动人体的轮廓宽度特征,提出了一种新的步态识别算法。该算法首先对每个序列进行运动轮廓抽取,同时从3个方向(水平、垂直、斜向)对时变的2维轮廓进行投影扫描,并分别转换为对应的特征向量;然后通过对级联的特征向量进行离散正交小波变换来提取低维步态特征,并抑制噪声;在此基础上采用支持向量机训练步态分类器组,最后用支持向量机组进行步态识别。在一组30人构成的步态数据库中进行的实验结果表明,该算法具备快速、稳健的特征,识别率达到91%,初步具备了实际应用的价值。

The automatic recognition by gait has recently gained more and more interests as the unique performance to recognition people at distance. An appearance-based approach to improve the gait recognition is proposed. The vector data scanned from horizontal, vertical and diagonal direction of the outer contour of binarized silhouette of a walking person are chosen as the image feature. These temporal and spatial feature sequences are decomposed based on the discrete wavelet transformation（DWT） to reduce data dimensionality and filter the noise produced from the procedure of template extracting. Then the multi-class support vector machine（SVM） models are trained by the decomposed feature vectors. The gaits are classified by the trained SVM models. This algorithm is applied to a data-set including thirty individuals. Extensive experimental results demonstrate that the proposed algorithm performs at an encouraging recognition rate with 91% at relatively lower computational cost.