A Classification Method for Judging the Depth of Chest Compression Based on CNN

The displacement can be calculated based on the integrated value of the acceleration signal waveform obtained by the acceleration sensor or gyroscope. However, this method is not effective in accurate measurement. Although some studies have improved the method of calculating accurate distance values by overcoming the effects of sensor noise or integration delay, the evaluation is still affected by sensor accuracy and environment. However, there are some special displacements, such as the chest compression. The displacement is a reciprocating motion and will return to the starting point again. Therefore, the acceleration waveform changes have obvious characteristics in the two stages from moving to the equilibrium position and returning to the starting point. Therefore, we propose an embedded classification method based on one-dimensional Convolutional Neural Network (CNN), which directly learns from the data of chest compressions and performs the signal formed by the Classification, distinguish the signal waveform under the standard pressing distance, so as to replace the calculation of distance measurement, and is not affected by factors such as pressure occlusion and electromagnetic wave interference, and has certain practical value on site. We tagged 937 presses and collected data from the simulator. The experiment evaluates the proposed network structure, and compares the classification results of the sample data with several CNN networks and Support Vector Machines with different structures in the literature. The results show that with sufficient training, the proposed one dimension CNN method can achieve an accuracy rate of more than 95%, and balances the accuracy rate and the hardware requirements.