Real-Time Chest Compression Depth, Rate, Posture, and Person Detections Using Binocular Vision for CPR Training

High-quality chest compressions (CCs) are crucial for cardiopulmonary resuscitation (CPR). This article presents a vision-based real-time CC detection system that can detect the CC depth (CCD), CC rate (CCR), CC posture (CCP), and CC person based on binocular vision and human key point (HKP) detection techniques. We propose a method combining target detection and color region segmentation to extract matching points during CC action identification. This method can achieve effective target ranging by detecting the motion trajectory of a MARK worn on the rescuer’s wrist. In addition, we propose a method to compensate for the CCD detection error caused by the rescuer’s palm deformation and arm tilt. We conducted volunteer CC experiments to verify the feasibility and effectiveness of our methods. The results showed that the mean absolute error (MAE) of the maximum CCD (MCCD) decreased from 14.306 to −1.073 mm after error compensation, with an average variance of 3.596 mm2. The MAE of the CCR was 2.4959 times/min. The recognition rate of the four types of CCP was 92.32%. The CC person detection accuracy was 92.9%. Our CC assessment method provides a noncontact measurement solution without installing any tool between the rescuer’s hand and the patient’s chest, which is convenient and has high application value in medical human-computer interaction.