Increasing Heart Rate Estimation Accuracy of Wrist PPG with EMG-Based Tissue Movement Reference

In optical heart rate monitoring, tissue movement artifacts, which are also known as micromotion artifacts, are caused by the movement of the tissue under the photoplethysmogram (PPG) sensor. These movements are originated from muscle activations, which causes changes in the optical path the light takes and in the skin-sensor contact force. These movements can be so small, that accelerometers cannot be used to detect them reliably. The use of electromyogram (EMG) as wrist tissue movement reference was investigated. EMG measures the electrical activity of the muscles and thus is directly connected to the movement of the muscles and surrounding tissues. Measuring EMG from a wrist is challenging due to the lack of muscle mass, causing low EMG Signal-to-Noise Ratio (SNR). A frequency domain spectrum subtraction algorithm was used to test the new movement reference signal. Weighted Frequency spectrums of EMG and acceleration signals were subtracted from the PPG spectrum, and peak detection and selection were used to choose the most probable heart rate related frequency peak. Comparison between the solution with both EMG and acceleration references and solution with only either acceleration or EMG reference was performed. The solution was evaluated by estimating the heart rate of 9 people when they performed seven-movement tasks divided in two sets. The first set consisted of four hand movements: wrist flexion/extension, hand squeezing, hand opening/closing, and individual finger movements. In the second movement set the subjects performed weight lifting with their bicep muscle, running, and running with a floorball racket in the left hand. All tasks were performed as repetitive movements. Results indicate that wrist-EMG can be used as a tissue movement reference for PPG artifact removal. When compared with only using the acceleration as a reference, the combination of EMG and acceleration references decreased the estimation error by 49% on average. The use of EMG enables robust removal of muscle artifacts, especially in situations where the tissue movement frequency differs from the dominant acceleration frequency. This could enable more accurate HR estimation during activities in which hand and forearm muscles are activated, such as biking, racket sports, and wall climbing