Real-Time Pig Vocal Analysis for Early Respiratory Infection Study

Pig farming has steadily advanced to rank among the top agricultural efforts. Their health and nutrition determine financial success in pig farming for weaning and pork production. Respiratory diseases are a significant cause of economic losses and animal welfare concerns in the pig industry. The ability to quickly intervene, improve treatment outcomes, lessen disease effects, and support long-term pig production depends on the early detection and diagnosis of respiratory infections in pigs of industrial farms. Vocalizations made by sick pigs may be an indication of respiratory illness. The proposed work uses the best performing MFE features trained model deployed to Arduino Nano, for real-time classification of pig vocals. The vocals are preprocessed using three signal processing algorithms: MFE, MFCC, and spectrogram. This research uses an Arduino Nano 33 BLE sensing development board and a mobile phone. The Edge Impulse studio receives the pig’s cough samples from Arduino, and the Mel cepstral features are extracted from the pig's sound signals and fed to neural network classifiers for training the CNN model. The model that used the MFE features extracted from pig vocals generated the best results, with a maximum accuracy of 99%.