Research On The Method Of Judging The Standardization Of Chest Compression And Error Compensation Based On Neural Network

In recent years,people often encounter various emergencies,leading to sudden cardiac arrest and sudden death.According to incomplete statistics,about 6 million people die of sudden cardiac arrest worldwide each year,and the number of sudden deaths due to sudden cardiac arrest in China is nearly 600,000.Chest compression as a basic emergency rescue measure requires that the rescuer's movements have high accuracy to obtain a good first aid effect,and at the same time,secondary injuries caused to the patient due to irregular compression should be avoided.The traditional method of judging whether the chest compression is standard or not is mainly realized in the dynamic process of measuring the compression depth.Current methods for accurately measuring depth include ultrasonic ranging,laser ranging and infrared ranging.Studies have shown that the use of micrometers to measure the depth of compression,using the quadratic integration method of the acceleration sensor value is effective.Combining data collection and experimental results,When an object reciprocates regularly,regular acceleration waveform data will be collected.When the action of chest compressions is standardized and meets the first aid standard,the two waveforms are more similar;When the error of the pressing distance difference exceeds 5mm,Compare the waveform with the standard distance waveform,there are more obvious differences;When there is large jitter during pressing or the interval between pressing is too long,there are more obvious differences too.This shows that the recognition and classification process of the waveform signal has learnable characteristics,and can be learned and classified using artificial intelligence methods,and the distance of the reciprocating motion of the object is evaluated according to the classification situation.The main work and innovation of this paper include:1.This study was fully considered,In the actual pressing process,need to continuously press the patient multiple times in a short time,In this way,there will be problems such as ranging occlusion,jitter,etc.This will make it difficult to obtain a large number of high-confidence data tags.For this problem,a weak supervised learning strategy based on weighted variance is proposed.Soft Max classifier can be used to construct a one-dimensional convolutional neural network to classify the compression waveform,and Adam optimization algorithm,Dropout and L2 regularization,learning rate attenuation and other methods are used to optimize the model,so as to make an intelligent judgment on whether the action standard of chest external cardiac compression is correct or not.2.Whether the traditional chest compression is standard,most of the quadratic integration method based on acceleration or its improvement.This study will analyze the causes of integration errors from the perspective of integration calculations.The reasonable basis of integral error compensation is given.Mainly based on the sampling frequency,Using three successive compressions of accelerated sampling data,and using the method of median calculation and median comparison to analyze the concavity of the waveform curve,to further determine whether the sampling segment is a positive error or a negative error.The concrete error compensation basis is given through theoretical derivation.Based on the traditional calculation of the pressing depth based on the accelerated quadratic integral,the method is optimized to improve the accuracy of the standard judgment of chest compressions.