Research On The Correlation Analysis Of Human Gait And ECG Signal In Motion Mode

Gait is the periodic and orderly movement with the time flies when people move,which has abundant information about the features of human motion behavior.ECG signals is a bioelectric signal produced by the activity of human heart,reflecting the physiological state of human health.The gait and the ECG signals are often widely used in medical diagnosis,medical treatment,sport science and health monitoring.As an important pattern of human body's gait features and ECG signals,the correlation studies between them is particularly important,especially provides important value and foundation for the prevention of disease and the assurance of health in the sport field.However,the correlation study between the gait and the ECG signals is in the theoretical stage,which lacks the quantitative analysis.In this paper,the gait features and the ECG signals in different motion modes was studied,the regression model was established and the quantitative correlation was analyzed.The main research are as follows:(1)According to the human pose data collected under different motion modes,the gait features was extracted with each ten gait cycles as a subset.First,the gait cycle,step length,step height,step width,step acceleration and maximum toe/foot heel angle were extracted by the quantization of gait temporo-spatial and angle.Then,approximate entropy,sample entropy,fuzzy entropy,lempel-ziv complexity and C0 complexity were extracted based on nonlinear dynamics method.Next,the value feature of wavelet coefficient mode was extracted based on wavelet transform method.Finally,multi-gait features(MGF)containing the time-and frequency-domain gait information are presented for analysis.(2)A real-time detection algorithm based on adaptive double threshold is adopted for the interference of various noises in the moving ECG signals.the noise and distinguishing QRS feature waves are first removed by bandpass filter,differential operation,nonlinear amplification and preprocessing of moving window integration.Then the R wave position is detected by double threshold adaptive optimization algorithm,and the missing detection and false detection of R waves are prevented by backtracking search and refractory detection.this algorithm improves detection speed and accuracy,thus locating R wave position in real time and accurately.finally,the ECG RR intervals feature corresponding to gait time are extracted for analysis.(3)The correlation analysis regression model based on BP neural network,random forest,Gaussian process regression,regular extreme learning machine and kernel extreme learning machine(KELM)between the gait features of human and the RR intervals of ECG prediction was studied.Besides,the accuracy and speed of ECG RR intervals prediction were compared experimentally with the evaluation indexes such as root mean square error,coefficient of determination and training time.The experimental results show that the lowest error of correlation prediction and the highest coefficient of determination can be obtained by the regression model based on KELM.The research found that the faster the gait speed of movement is,the more accurate the RR intervals of ECG signals is predicted.At the same time,the MGF is superior to a single feature and highly correlated with the ECG RR intervals.