基于声音特征优化和改进支持向量机的鸟声识别

为了在低参数量下提高鸟鸣声的识别准确率，提出了一种新的鸟声识别方法，包括鸟声信号特征优化和乌鸦搜索-支持向量机（Support Vector Machine,SVM）分类识别。该方法首先采用主成分分析法对从鸟声中提取的梅尔频率倒谱系数（Mel Frequency Cepstrum Coefficient,MFCC）和翻转梅尔频率倒谱系数进行选择，得到优化后的声音特征参数并将其作为鸟声识别算法的输入；然后利用乌鸦搜索算法对SVM的核参数和损失值进行选优，得到改进的SVM网络用于鸟声分类识别。试验结果表明，该方法对5种鸟声识别的准确率为92.2%，声音特征维数在16时可以得到最好的识别效果。该方法为野外鸟声自动识别提供了一种可行的方式。

Bird Sound Recognition Based on Optimized Sound Features and Improved SVM

To improve the accuracy of bird sound recognition with low number of parameters,a new bird sound recognition method is proposed,including optimization of bird sound signal features and crow search support vector machine (SVM) classification recognition.Firstly,principal component analysis is used to optimize the Mel frequency cepstral coefficients (MFCC) and flipped Mel frequency cepstral coefficients extracted from bird sound,and the optimized sound features parameters is taken as input for the bird sound recognition algorithm.Then,the crow search algorithm is used to optimize the kernel parameters and loss values of the SVM,and an improved SVM network is obtained for bird sound classification and recognition.The experimental test results show that the correct recognition rate of the method for five bird sounds is 92.2%,and the best recognition effect can be achieved when the sound feature dimension is 16.The method provides a feasible approach for automatic bird sound recognition in the wild.