Feature extraction for poultry vocalization recognition based on improved MFCC

Animal’s sound has rich information. Animal’s health status and adaption for environment can be fed back from vocalism. Vocalization has been the important way for measuring animal welfares. For the auto recording and analysis of animal sound, some ways were studied for sound feature extraction. On the basis of Liner Prediction Coefficient (LPC) and Mel-Frequency-Cepstral Coefficient (MFCC), a new algorithm for feature extraction－Average MFCC was introduced. For evaluating the improvement of poultry sound recognition, SVM classification model was used to perform the experiments, the results showed that performance was obviously improved.