The Use of Wrist EMG Increases the PPG Heart Rate Accuracy in Smartwatches

Abstract

The impact of tissue movements on the accuracy of heart rate (HR) estimates is a challenge in today’s wearable technology. Tissue movements are caused by muscle activity that modifies the optical path of the reflectance photoplethysmography (PPG), leading to motion artifacts (MAs) that mask the true HR. This kind of MA is not always detected using accelerometers (ACCs). In this study, we propose a method to increase the PPG HR accuracy of a wristwatch using wrist surface electromyogram (EMG) and ACC using spectrum subtraction algorithms. We collected the wrist EMG, wristwatch PPG, ACC data, and the electrocardiogram (ECG) from nine subjects. Data were recorded during four frequent hand movements and three activities (weightlifting and running/walking with and without holding a racket). The added value of the EMG was studied. Visual results indicate that wrist EMG correlates well with the MA seen in the PPG signal and provides additional information over the typically used ACC data. Our analysis showed that the proposed artifact removal method using EMG and ACC decreases the HR estimation error on average by 49% compared to only ACC. Our results showed that wrist EMG contains complementary information on the PPG artifacts and offers a novel signal modality for improving optical HR estimation accuracy in smartwatches.