Monitoring pig respiration frequency using Wi-Fi wireless sensing technology

Abstract: Monitoring pig respiration timely in swine farms is critical to safeguard swine production. Traditional manual method by tagging pigs with sensors is inefficient and makes pigs stressful. Figuring out non-contact and non-destructive ways is hence necessary. Wi-Fi technology is non-intrusive and robust, and it has received increasing attention over the past few years as a potential method to track animal respiration. Its fundamental principle is that the exhaling - inhaling cycle in pig respiration results in a small change in the Wi-Fi signals when they propagate from transmitter to receiver. In the 802.11 a/g/n standard, the signal response in the channel can be partially extracted from the off-the-shelf OFDM receivers in the format of Channel State Information, which revealed that a set of channel measurements can indeed pick up such change, making it feasible to monitor animal respiration. We proposed a novel method based on the Wi-Fi signal in this paper to estimate the respiration rate of pigs reared in a single shed. We obtained the motion-state data in the CSI data files first using the off-the-shelf Wi-Fi devices commonly used in daily life. The CSI data is matrix of 1×3×30, where 1 is the number of transmit antennas, 3 is the number of receive antennas and 30 is the number of subcarriers in one beam. Preprocessing these data and evaluating the carrier periodicity level enabled us to identify the CSI signal sequences that contain the abdomen undulation of pigs. This is followed by smoothing the subcarrier curve with the algorithm of smoothing spline and evaluating the period and frequency of the subcarrier with the self-correlation function of CSI sequences. Finally, we statistically estimated the weighting average of the multiple subcarrier frequencies to calculate the respiration rate of pig. Taking the number of breaths manually accounted per minute from the pigs as ground truth, the proposed method was tested against the respiration data measured from 9 piglets, 5 fattening pigs, 3 pregnant sows and 3 sick pigs with abdominal breathing caused by illness. The results show that the maximum relative error is 3.18%, and the average relative error is 1.398%. The study has wide implications in using Wi-Fi technology to monitor respiration of animal.