Research On Algorithms Of Wearable Physiological Signal Monitoring Device For Smart Phone Applications

Wearable physiological signal monitoring is an important technique in the field of medical monitoring.Continuous dynamic electrocardiogram（ECG）and respiratory signal monitoring can provide early warning and diseases risk assessment for cardiovascular diseases and respiratory diseases.Unfortunately,the continuous ECG and respiratory signals collected by wearable devices are susceptible to interference from environmental noise and body movement,resulting in uneven signal quality.However,discarding the interfered signal will increase the workload for repeated signal collection,while analyzing the signal without processing will increase the burden of medical staff and reduce the accuracy of diagnosis or even cause misdiagnosis.Therefore,in order to effectively utilize the physiological signals collected by wearable devices,it is very important to propose signal processing algorithms,as well as robust methods for parameter extraction with stronger anti-interference ability.In this paper,we proposed a heart rate detection algorithm and a signal quality assessment algorithm in order to improve the anti-interference ability of existing algorithms and improve the multi-classification accuracy of ECG quality assessment.For the acceleration（ACC）signals,we proposed a human activity recognition algorithm.With the assistance of the ACC signal,we proposed an ECG motion artifacts suppression algorithm and a respiratory rate detection algorithm.In addition,a mobile phone application was designed to display physiological signal waveforms and heart rate for monitoring visualization.Three major works have been done in this paper:1.Collection and annotation of physiological signals in different motion statesIn this paper,a wearable monitoring device was used to collect the ECG and ACC signals of 29 healthy male volunteers at rest and in different motion states.The collected signals were intercepted into 4s segments,and a total of 7133 data segments were generated,and the ACC and ECG signals were annotated simultaneously.Finally,a human activity recognition dataset（7133）,a heart rate dataset（7016）and a signal quality assessment dataset（7133）were generated.The human activity recognition dataset included resting,walking slowly,walking fast,jogging and fast running states.The signal quality assessment dataset contained three quality levels（good,acceptable and unacceptable ECG）.There was a certain correlation（R=0.61）between the motion state and the quality of the ECG signal,and the motion state could be used as an auxiliary index to judge the quality of the ECG signal.2.Design of the algorithmsFirstly,the human activity recognition algorithm based on ACC signal was developed.The mean,skewness,peak frequency,interquartile range and the time domain integral were used as the classification indicator.The support vector machine was used to automatically recognize the states of the human.The result showed that the accuracy was 97.62%.Secondly,a heart rate detection algorithm was designed.The ECG signal was preprocessed to remove the baseline drift and the high-frequency interference,the signal was coarse-grained by adaptive peak dilation and waveform reconstruction,heart rate was calculated based on the frequency spectrum obtained from fast Fourier transformation.In order to avoid the influence of motion interference on the accuracy of heart rate detection,an anti-interference strategy based on high-amplitude interference suppression was added.This strategy located and suppressed motion-related high-amplitude interference by presetting a threshold and compressing signals.The results showed that the correlation coefficient between the calculated heart rate and the standard heart rate was 0.999.The accuracy of the proposed algorithm was significantly higher than the wavelet transform method in all states,including resting（99.94%vs.99.10%,P<0.01）,walking（100%vs.97.25%,P<0.01）and running（100%vs.90.89%,P<0.01）.The absolute error[0（0,1）vs.1（0,1）,P<0.01]and relative error[0（0,0.59）vs.0.52（0,0.72）,P<0.01]of the proposed algorithm were significantly lower than the wavelet transform method during running state.In addition,according to the characteristics of continuous heart rate changes,this paper also proposed an anti-interference strategy based on time sequence characteristics to achieve real-time continuous heart rate monitoring during exercise.Thirdly,an ECG signal quality assessment algorithm was designed.The algorithm proposed 5 effective indicators,including indicators based on multiscale nonlinear amplitude distribution（adSQI₁,adSQI₂）,indicator based on energy ratio（ptSQI）and indicators based on heart rate（tHR,rHR）.The back propagation neural network combined these indicators to divide the signal into ECGs of good quality,acceptable and unacceptable.In order to verify the effectiveness of the algorithm,we proposed a signal quality assessment-based ECG wave delineation algorithm,and detected R or T waves according to the classification results.The results showed that the classification accuracy of this algorithm was 96.74%,and the detection accuracy of R wave and T wave were 99.95%and 99.57%,respectively.This algorithm improved the signal utilization rate while ensuring the accuracy of diagnosis.Fourthly,a recursive least square notch filter was designed in combination with the ACC signal to suppress the ECG motion artifacts.The results showed that the ad SQI₁ after filtering was increased by 21.73%compared with that before filtering.The signal filtering method could effectively suppress the ECG motion artifacts and improved the signal quality.Finally,a respiratory rate detection algorithm was designed.The algorithm extracted the respiratory rate in the frequency domain and used the ACC as an auxiliary signal.The results showed that the accuracy was 91.84%,and the maximum absolute error was within 6time/min,which could realize the detection of respiratory rate under low-intensity activities.3.Design of the APP based on smart phoneAn application（APP）software that can display physiological signal waveforms and parameters in real time was designed.This paper had completed the user interface design,realized Bluetooth communication,and displayed the real-time signal and critical physiological parameters.The proposed human activity recognition algorithm,anti-motion interference heart rate detection algorithm,ECG signal quality assessment algorithm,ECG motion artifacts suppression algorithm and respiratory rate detection algorithm ensure accurate analysis of physiological parameters under exercise.The APP software based on the Android system allows users to check their physiological signals and parameters at any time.The entire system is potentially to be used in physiological wearable monitoring,and to provide technical support for health monitoring and early diagnosis and early warning of diseases.