HRCal: an effective calibration system for heart rate detection during exercising

Heart rate directly reflects heart health and the detection of heart rate contributes to finding the abnormal performance of heart activity in a timely manner. Nevertheless, there is scope for a significant improvement in current heart rate detection systems and devices, especially during strenuous exercise. Motion compensation algorithm is used in most current systems to improve the monitoring accuracy, but it is limited by sensors and its performance is not satisfactory. In this paper, we propose HRCal, a novel Heart Rate Calibration System, which establishes a Long Short-Term Memory (LSTM) model to calibrate the detection of heart rate based on multisensor data fusion. Specifically, HRCal utilizes the built-in sensors (e.g. accelerometer, gyroscope and magnetometer) from smart devices (smartphones and sports watches) to collect users' motion data. Then a LSTM model is proposed and trained with different features to improve the accuracy and reliability of heart rate detection. In addition, we also elaborately design an evaluation scheme to compare HRCal with other approaches. We have fully implemented HRCal on Android platform and the experimental results (8 subjects) demonstrate that HRCal has a remarkable effect on common sports watches, to improve their accuracy of heart rate detection in physical training (up to 12.5% for moto 360 and 6.8% for Mio Alpha).