Research on Bird Songs Recognition Based on MFCC-HMM

Abstract

HMM (Hidden Markov Model) is a statistical-signal based model, and MFCC (Mel frequency Cepstrum Coefficient) is one kind of characteristic parameters, both of which are widely used in speech recognition. This paper studies MFCC-HMM bird songs recognition technology. Firstly, it collected bird songs through web crawler. Then, bird audios were preprocessed and MFCC features were extracted. Through differential methods calculation, improved MFCC feature parameters are obtained, which are fed into the HMM model to classify different kinds of bird songs. Experimental results show that the MFCC-HMM model achieve a recognition rate of 90.47% among the six kinds of birds.