基于广义回归神经网络的小麦碰撞声信号分类

Wheat Kernels Impact Acoustic Signal Classification Based on GRNN Neural Network

小麦在储藏阶段由于各种灾害导致损失巨大,并降低了面粉质量,及时检测并分离小麦的受损颗粒迫在眉睫。文章以提取4类小麦碰撞声信号为基础,使用数字信号处理方法对小麦完好粒、虫害粒、霉变粒及发芽粒的碰撞声信号提取有效特征,最后利用广义回归神经网络进行分类,对于3类小麦类型的识别取得了较好的识别率。应用结果表明广义回归神经网络能够较好地实现区分受损小麦颗粒与完好小麦颗粒。

In storage stage,a huge number of wheat will loss due to various disasters and the damaged wheat kernels reduce the quality of the flour,it is important to detect and separate the damaged wheat kernels timely. This paper uses digital signal processing method to analyze the impact acoustic signal of un-damaged kernels,IDK(Insect Damaged Kernels),slab-damaged kernels and sprout kernels,some characteristic features of the four types wheat kernels are extracted and the wheat kernels with GRNN Neural Network are classified,a good recognition result of three types wheat kernels in the end is gotten. The research shows that the GRNN Neural Network is useful to the detection and separation of damaged wheat kernels and un-damaged wheat kernels.