Analysis And Implementation Of Swimming Monitor System Based On Triaxial Accelerometer

Today's people's lifestyle and work pressure make most of them in a sub-health state.Sports monitoring system can prompt people to record their daily exercise and data,helping form a good exercise habit and lifestyle.With the popularity of smart watches and technological development,more and more smart watches have a high level of waterproof capabilities,with accelerometer embedded.Swimming monitoring based on accelerometer is a new research direction.As a popular sport,the swimming monitoring system can help people to monitor and record their own swimming data easily and formulate a reasonable training plan.The main content of this article is analysis and implementation of the swimming s monitoring system based on the accelerometer.The main tasks include:1.Use embedded accelerometer in android device to collect acceleration data.and propose preprocessing methods such as data filtering and windows dividing.Aiming at instability of sensor's working frequency,a re-sampling method is proposed to ensure that the data frequency is around 50.Low-pass filter is used to filter initial data,which is used in the system to identify the stroke style,count strokes,identify turning behavior.2.For strokes style recognition,two new data features,peak-to-peak distance and peak-to-valley distance,are proposed,which enrich the feature set and form a 24-dimensional feature set with other time domain features,which are used to train different classification models separately.The SVM classifier has the best recognition accuracy,88.46%,to distinguish four normal strokes.Finally,comparing the influence of different time windows on recognition accuracy,we find that the optimal window size is 3 seconds.3.Aiming at the detection of strokes based on accelerometer,a region peak detection algorithm is proposed.Compared with the dynamic threshold method and one-peak detection method,new method has a higher detection accuracy of 93.57%.Besides,a single axis is used to identify turning behavior with threshold method.By combining with strokes amount,additional swimming data is formed to enrich the swimming monitoring results.4.Design and implement a swimming monitoring system based on accelerometer i n C/S framework,including modules of data collection,preprocessing,feature extraction,classification and identification,to monitor data of swimming time,distance,strokes,calories,stroke rate,DPS,time spent per lap,etc.Verify its effectiveness through functional tests and accuracy tests.In actual using,the accuracy of strokes style recognition declines.Thus,the self-data learning function is proposed to solve the effect of differences between individual bodies' movements.The average accuracy of the strokes style recognition in the case of individual use is 95.49%.And the improved state correction and results collection improve applicability of the system and rationality of the results.