Recognition of sick pig cough sounds based on convolutional neural network in field situations

Abstract

Coughing is an obvious respiratory disease symptom, which affects the airways and lungs of pigs. In pig houses, continuous online monitoring of cough sounds can be used to build an intelligent alarm system for disease early detection. Owing to complicated interferences in piggery, recognition of pig cough sound becomes difficult. Although a lot of algorithms have been proposed to recognize the pig cough sounds, the recognition accuracy in field situations still needs enhancement. The purpose of this research is to provide a highly accurate pig cough recognition method for the respiratory disease alarm system. We propose a classification algorithm based on the fine-tuned AlexNet model and feature of the spectrogram. With the advantages of the convolutional neural network in image recognition, the sound signals are converted into spectrogram images for recognition, to enhance the accuracy. We compare the proposed algorithm’s performance with the probabilistic neural network classifier and some existing algorithms. The results reveal that the proposed algorithm significantly outperforms the other algorithms—cough and overall recognition accuracies reach to 96.8% and 95.4%, respectively, with 96.2% F1-score achieved.