Abstract
It is important to prevent infectious disease expansion in swine by detecting respiratory tract infections in early stage. Counting sneezing and cough is conventionally used to detect the disease. In this paper, we propose the automatic monitoring system of swine sneezing and cough using a microphone-array, which achieves an automated recognition and localization procedure for swine in weaning period cough and sneezing sounds. To realize this system, it must be achieved that a swine cough and sneezing sound are detected successfully in spite of many noises in swine house. The noises cause detection errors in the system. A technique using a high-pass filter has been found effective to detect the sneezing that includes frequency component of over 30 kHz frequency components than noises. Another technique using a band-pass filter has been found effective to detect the cough that includes frequency component in the range of 10 – 20 (kHz). These detection techniques and a localization technique based on the time difference of arrival are evaluated in experiment. The obtained results suggest that the proposed technique worked with sufficient accuracy. The correct discrimination rate for sneezing was 99.9 % and correct location estimation rate for detected sneezing was 85.7 %. The correct discrimination rate of cough was 99.6 % and correct location estimation rate for detected cough was 75.0 %. These accuracies were enough to achieve automated recognition and localization procedure for swine cough and sneezing sounds.
