Identification Of Abnormal Calls In White-feathered Broilers Based On Sound Detection

With the rapid development of China's economy and the continuous improvement of the consumption level of the whole society,people's demand for livestock meat,especially poultry meat products is also growing.Among them,in respect of meat industry structure,broiler chicken accounts for a large part of proportion.However,in the process of raising broilers,breeders cannot ensure the health of broilers.As sound signals are easy to obtain and process and contain a large amount of information,it is possible to monitor and predict the health of broilers through sound.In this paper,an audio acquisition system was first designed to collect and transmit the sound signals of white-feathered broilers,and then several filtering methods were used to process the collected sound signals.Through comparative analysis,the Wiener filter was used to de-noise the sound signals,and the sound signals with high signal-to-noise ratio were obtained.In the process of feature extraction,the three parts of time domain and frequency domain,MFCC and sparse representation are extracted respectively,and a total of 60-dimensional features are obtained.Considering the complexity of calculation,after normalization processing,random forest is used to screen the features,and finally 30-dimensional features are obtained.Then,the random forest was used as a classification model,and the inherent parameters of the random forest were optimized by combining grid search algorithm and cross validation,so as to realize the classification of the sound signals of white feathered broilers and improve the recognition accuracy of each sound signal.Finally,a visualization platform is built to analyze and display the predicted sound signals.When the proportion of cough sound in a certain unknown sound signal exceeds the set warning threshold,a prompt box will pop up on the software platform of the system to prompt the warning information.The research in this paper effectively realized the problem of voice signal recognition of white-feathered broilers and improved the accuracy of voice signal recognition of white-feathered broilers.According to the data statistics of abnormal sound of broilers,the health status of broilers can be judged,so it has certain practical value and good application prospect.