飞行状态下四种菊头蝠回声定位声波的小波包识别方法

RECOGNITION METHOD BASED ON WAVELET PACKETS FOR ECHOLOCATION CALLS FROM FOUR SPECIES OF Rhinolophus IN FLIGHT

研究了飞行状态下的四种菊头蝠回声定位声波的识别方法。通过小波包分解得到各个频带能量作为识别特征向量,用主成分分析法优化特征空间。提取少数几个主成分,这些主成分彼此不相关,符合特征优化的要求,以主成分向量作为BP神经网络的输入对蝙蝠的种类进行识别。个体识别正确率达到了80%以上,表明基于小波包分解和神经网络识别的方法对蝙蝠回声定位声波进行识别是可行的。

The identification method of the echolocation calls from four species of Rhinolophus in flight was studied based on the wavelet packet decomposition. The energy values of different frequency channels of sound signal were extracted as feature vector by wavelet packets decomposition. Then feature vector was optimized by principal components analysis. A few principal components, which were irrelated with each other and in accordance with the demand of feature optimization, were used as the inputs of BP neural network to identify the bats. The identification correct rates for every species could come up to over 80% simultaneously. The result shows that this method is feasible to recognize the different species of Rhinolophus in flight.