基于HHT方法的果蝇鸣声特征提取及分类

Feature extraction and classification of fruit fly's flight sound based on HHT

采用HHT方法对同种内2个不同品系果蝇翅振鸣声进行特征分析,分别提取果蝇翅振鸣声前10阶IMF能量与信号总能量的比值,HH谱图的低频段、中频段、高频段的相对能量值作为特征向量.设计BP神经网络分类器识别不同品系果蝇.实验结果表明,用HHT方法提取特征,神经网络识别不同品系果蝇的方法是可行而有效的,为进一步鉴别果蝇种内关系提供了新的思想和方法.

Using HHT method,the paper analyzed the wing vibration sound of two different strain of fruit flies in the same species.It extracted effective characteristics of fruit fly＇s wing vibration sound,which contained ratios of first ten IMF and signal total energy,the relative energies of low-frequency stage,middle-frequency stage and high-frequency stage in HH spectrum.Then,the paper designed BP neural network to identify different strain of fruit flies.The experiment result indicated that it was feasible and effectivethe method of extracting features by HHT and identifying different strain of fruit flies by neural network.And it provided a new idea and technique for further discrimination of intraspecies relationship of fruit fly.