Remote Heart Rate Estimation of Canines Using a Mmwave Radar and Depth Camera

While heart rate estimation with frequency modulated continuous wave (FMCW) radar on humans has become a common research topic, the potential of this technology for vital sign monitoring of animals, such as canines, has had little exploration. This letter serves as a study on the feasibility of applying mm-wave FMCW heart rate estimation technology on conscious canines. A data capture system and remote heart rate estimation pipeline is implemented and used on canine subjects. Several methods are compared which include peak counting, spectrogram analysis, long-short term memory (LSTM) neural networks and a hybrid digital signals processing (DSP)-LSTM approach. While the dogs were not stationary enough for traditional DSP approaches to produce meaningful results, the hyrbid DSP-LSTM approach improved the overall accuracy (13.9 root-mean-square (RMS) error bpm and a 0.76 correlation coefficient). These improvements will greatly increase the feasibility of mm-wave technology for future applications in livestock monitoring and wildlife conservation.