Abstract:Respiratory tract diseases,frequently outbreaking in pig barns,bring economy lost in modern pig production.In this paper,a method was proposed to predict and monitor respiratory tract diseases by recognition of the cough sound of pigs.The sounds of pig were processed by the spectral subtraction denoising and the endpoint detection before being made into the improved Mel-frequency cepstral coefficients (MFCC),which are feature vectors input into the vector quantization matching algorithm (VQ algorithm).Thus a pig cough recognition model based on the VQ algorithm was developed to recognize the sound of cough.The results showed that the recognition rate,the false positive rate and the comprehensive recognition rate of the model based on the improved MFCC were 91%,12% and 90.0%,respectively,which were better than 88%,14% and 87.3% of the standard MFCC,suggesting the feasibility of the early warning of the respiratory tract disease of pigs in the future.