基于小波分析法的传感器信息融合的运动损伤检测方法

传统方法针对多组传感器路径中的检测点,在很大程度上会出现若干存在差异的损伤发生概率,导致运动损伤检测不准确。为此,提出一种传感器信息融合的运动损伤检测方法。利用多帧帧间差的累积消除空洞效应,在此基础之上,融合传感器确定出准确的人体运动区域,以此对不同场景人体运动进行监测;采用小波分析法对监测结果的非平稳信号进行分析,得到运动损伤特征。将传感器信息融合和小波神经网络结合在一起,获取所有传感器的小波能量特征向量,按照最大概率密度函数值和特征向量获取融合运动损伤检测结果及损伤种类。实验结果表明,所提方法检测结果准确,实用性强。

Motion damage detection method based on wavelet analysis for sensor information fusion

In view of the detection points in the multi group sensor path, the traditional method, to a large extent, appeared a number of differences in the probability of damage, which led to the inaccurate detection of motion damage. To solve this problem, a motion damage detection method based on sensor information fusion is proposed. The accumulation of multi frame inter frame differences was completed to eliminate void effect. The motion area of human body, which was obtained by a single sensor, contains noise, and the texture is insufficient. There were still holes in the body. The human motion areas extracted by multi sensors were fused together to determine the accurate human motion area. The human motion was monitored in different scenes. The non-stationary signals were analyzed by wavelet analysis, and the characteristics of motion damage were obtained. Combining the sensor information fusion with the wavelet neural network, the wavelet energy eigenvectors of all sensors were obtained. The obtained wavelet feature vectors were fused together. As input of the wavelet neural network, the neural network training and fusion calculation wre completed, and the fusion motion damage was obtained according to the most approximate rate density function value. Test results and types of damage. Experimental results show that the proposed method is accurate and practical.