中文摘要：

研究了一种基于多级分类模型的非特定人走路模式识别算法，实现了对水平行走和上、下楼梯三种运动状态的识别．将装有微型加速度传感器的无线数据采集装置固定于人体后腰部，获取运动时的三维步态加速度信号．采用离散小波变换提取与运动相关频带的时频特征，并结合步频以及垂直方向和前进方向加速度信号之间的互相关性，经过特征融合设计了多级分类识别算法．通过对10个人共360组数据的测试结果表明：在步频范围扩大到1～3Hz时，识别率达到了96．1％，且对测试对象的依赖性小．

英文摘要：

This paper presented the multiple classifier based walking pattem recognition algorithm, which could identify three walking patterns： horizontal walking, up and down staircase walking. Three-dimensional accelerations during walking were acquired from the wireless accelerometer device fixed on the back waist. The discrete wavelet transformation was applied for time-frequency analysis. The time-frequency features associated with the main frequency band of the motion, walking cadence and the correlation between the vertical and forward acceleration signals were combined to design a multiple classifier. A set of 360 gait samples involving 10 people were used for test,giving an overall recognition accuracy for 96.1% when the walking cadence range was within 1 - 3Hz,and this algorithm was less dependent on individuals.