Monitoring of swine sneezing using time-frequency analysis for detecting diseases

Abstract

This research mentions a detection method of swine sneezing in noisy swine house. While it is important to prevent swine infectious disease expansion in early stage, it is difficult to detect an infection in early stage because we cannot judge an infection from appearance of swine. However, it is a well-known fact that a swine with an infection tends to sneeze much more than healthy swine. The aim of this research is to achieve a detection of diseased swine in an automated way using the sound in the swine house and signal processing. We found that the sound of swine sneeze includes high frequency components 5-30 kHz and a sound feature of spectra, compared to the environmental noise in swine house. Based on these features of sneezing sound, we propose the detecting technique of sneezing sound based on the composition template of a frequency feature by a time-frequency analysis which is calculated from a lot of sneezing sounds. This method calculates the difference between frequency values of templates and recorded sounds. That could judge the recorded sounds as sneezing if sounds are smaller than the threshold. As a result, the method demonstrated effectiveness in a detection of swine sneezing.