LAUSR

Traditional fitness equipment usually equips with a contact-type device to measure heart rate (HR), which may limit the user’s hand movements and expose the user to risks of contagious diseases. A camera-based method for non-contact HR measurement has been used to solve the above problems; however, the method has the problem of motion artifact to be dealt with. In this article, we proposed an adaptive skin color detection method and several new pulse rate measurement methods in the frequency-domain to increase the accuracy of pulse rate measurement based on the previously proposed chrominance-based adaptive filter and normalization (CADN) and domain selection scheme (DSS) method. The experimental results reveal that the mean absolute error and root mean square error (MAE/RMSE) of pulse rate measurement in real-time are 2.11/2.93, 2.43/3.44, and 2.26/3.45 beats per minute (bpm) for biking, stepping, and treadmill exercises, respectively. Compared with the previous CADN + DSS method, the proposed methods can improve the accuracy of pulse rate measurement.