立体栖架散养蛋鸡行为加速度信号降噪方法比较分析

Comparison of the denoising methods for the acceleration signals collected by laying hens’ behaviors monitoring equipment inside a 3D aviary system

针对采用姿态传感器监测散养蛋鸡行为过程中存在加速度信号易受噪声干扰,不同行为加速度信号特征尚不清晰的问题,以立体栖架散养蛋鸡的三维空间行为为目标,在构建蛋鸡穿戴式行为监测系统,获取不同行为加速度信号的基础上,研究小波阈值、CEEMDAN、CEEMDAN结合小波阈值对加速度信号的降噪效果,分析蛋鸡6种典型行为活动的加速度信号特征。结果表明:1)小波史坦无偏似然估计(Rigrsure)阈值的降噪效果更好,降噪后信号的信噪比(SNR)为17.613 2,均方根误差(RMSE)为0.045 0;2)根据X、Y轴加速度标准差分布特征,可将蛋鸡三维空间行为活动强度由低到高分为趴卧和站立、采食和饮水、行走、跳跃4类,各类行为活动的加速度信号标准差差异极显著(P<0.001)。本研究表明,小波Rigrsure阈值法在对蛋鸡空间行为活动加速度信号进行降噪的同时可以保留更多的有效信息,可根据加速度标准差对蛋鸡典型行为的活动水平进行分类识别。

Acceleration signals generated from wearable behavior monitoring systems are susceptible to noise interference.Moreover,the distribution characteristic of acceleration signals of various behaviors is not yet clear,which significantly restricts automatic behavior recognition.In this study,a wearable behavior monitoring system was constructed to collect behavioral acceleration signals of the laying hens inside the 3Dspace of an aviary system.The performances of three denoising methods on accelerations were compared,including wavelet threshold,CEEMDAN and CEEMDAN combined with wavelet threshold.Moreover,the acceleration signal characteristic of six typical behaviors of laying hens was analyzed.Results showed that:1)Wavelet Stein’s unbiased risk estimation(Rigrsure)threshold had better denoising performance.The signal to noise ratio(SNR)and root mean square error(RMSE)of the denoising signals were 17.613 2and 0.045 0,respectively.2)According to the distributions of standard deviation of X-and Y-axis accelerations,the activity levels were divided into four categories from low to high level,i.e.,lying and standing,eating and drinking,walking,jumping.The standard deviations of acceleration signals of various behavioral activities were significantly significant(P<0.001).This study indicated that the wavelet Rigrsure threshold method could retain more effective information while reducing noises,and the activity levels of typical behaviors of laying hens could be classified according to the standard deviation of acceleration.