人工神经网络对果蝇鸣声的分类识别

Classification of fruit fly's sound by artificial neural network

为了利用昆虫鸣声对昆虫进行种间或种下分类，对实验室环境下同种2个不同品系黑腹果蝇的飞行翅振鸣声进行了采集、分析，提取呜声信号特征参数，并利用人工神经网络对采集的果蝇鸣声信号进行分类识别。结果表明，2个品系果蝇鸣声的基频均为236．86Hz，有多个谐频，频率范围为O～4000Hz，重叠较大；所建立的人工神经网络对种内不同品系果蝇鸣声的正确识别率均在75％以上，识别效果很好。研究结果为果蝇种下分类提供了新的方法和依据。

In order to use sounds of insects to classify the interspecies or subspecies, this paper has col lected and analyzed the sound of two strains of Drosophila melanogaster, extracted their sound feature pa- rameters,and conducted the classification of the different strains of fruit fly＇s sound by using neural net- work. As the result of the experiment demonstrates,the fundamental frequencies of the strains’ sound are all 236.86 Hz,with many harmonics and frequency from 0 to 4 000 Hz and have overlaps in the frequency; the established neural network is effective in identifying the sounds of different strains of same species,and the average accuracy of identification is above 75 %. The result of this research provides a new way and ba- sis for subspecies classification of fruit fly.