Chicken Sound Recognition Using Anti-noise Mel Frequency Cepstral Coefficients

A new chicken sound recognition technology based on the AMFCC parameters is proposed to improve the accuracy of chicken sound recognition under different kinds of noise environments in the real world. First, the noise estimation algorithm for highly non-stationary environments is used to estimate the noise power spectrum of the chicken sound in the noise environment. Second, the multi-band spectral subtraction is taken over the noisy sound to achieve the clean sound spectrum. Third, the estimated clean chicken sound spectrum is taken as input for the process of MFCC extraction. Then, we combine every three continuous frames´ parameters with their perspective short-time energy to achieve the middle frame´s AMFCC. Finally, the experiments of chicken sounds recognition under different noise environments are constructed. The experimental results show that the AMFCC parameters can effectively weaken audio noise disturbance and improve the recognition rate compared with the traditional MFCC parameters and PNCC parameters in the low signal noise ratio.