小麦碰撞声信号时域建模与分类

分析小麦碰撞声信号,可识别受损小麦。提取三类小麦碰撞声信号,分析小麦碰撞声的时域特征,建立合适的拟合模型,并提取残差平方和、判定系数、峰值振幅等六个时域特征;最后利用BP神经网络进行分类,发现小麦完好粒、虫害粒及霉变粒碰撞声信号的时域特征存在差异,并取得了较好的识别率。应用结果表明选用适当的数学模型能够较好地拟合小麦碰撞声信号,实现区分受损小麦颗粒与完好小麦颗粒。

Time domain modeling and classification of wheat impact acoustic signal

Analyzing the wheat impact acoustic is help to recognize the damaged wheat kernels. This paper extracted three types of wheat impact acoustic signal, and analysed the time domain characteristics of all the acoustic, established a suitable fitting model, and got the statistical parameters of SSE, the peak amplitude and other four time domain characteristics, and used BP neural network to class the wheat kernels. The experiment finds that the time domain characteristics of undamaged wheat kernels, the insect damaged kernels and the moldy kernels are different, and have achieved a better recognition rate. The application results show that the selection of appropriate mathematical model can fit the wheat impact acoustic signal better, distinguish undamaged wheat kernels and damaged wheat kernels successfully.