摘要
以常见的 7 种飞虱雄虫求偶鸣声信号的频率峰值作为输入向量,用人工神经网络来识别它们的鸣声,平均识别率达 90.6%。人工神经网络可以用于昆虫鸣声识别。
Artificial neural network (ANN) could be applied in identification of insect acoustic signals. Here male courtship signals in seven species of planthoppers were recorded and analyzed. The results showed that their frequencies were overlapped. However, these signals could be identified by ANN and the average accuracy of identification was 90.6% if their peak frequencies were as input vectors. But they couldn’t be distinguished if their waveforms were as input characters. Therefore, selecting appropriate input characters were important when using ANN for identifying insect acoustical signals.
基金
国家自然科学基金项目(30070427)
植物病虫害生物学国家重点实验室开放基金资助
关键词
人工神经网络
飞虱
鸣声
识别
artificial neural network
planthopper
acoustical signal
identification
