关键词: 步态识别, 活动能量图, 分解向量, 快速傅里叶变换, 动态空间规整法

Abstract: This paper proposes a gait recognition algorithm based on mass energy image and spectrum analysis. The method includes following steps: firstly, two-value silhouette sequences extraction is performed for each subject; active energy image is calculated by adding up the difference of each image in a sequence; the row mass vector and column mass vector are extracted respectively. Secondly, the space mass energy image is spectrum analysed by windowed and Fourier transform techniques, combining the extreme-value computing method to construct gait feature vector. It uses dynamic space warping for classification under space mass energy image and spectrum analysis. Enough experiments are performed on CASIA Dataset B. The final average recognition rate reaches 93%, and with a robustness of backpacks and clothes. The method has been demonstrated higher recognition rate than other methods.

Key words: gait recognition, Active Energy Image（AEI）, mass vector, Fast Fourier Transform algorithm（FFT）, Dynamic Space Warping（DSW）