Research Of Gait Recognition Based On Surface Electromyography Signals

The mechanization degree of agriculture is improved with the enhancement to overall national strength and the advanced scientific and technological achievements.The miniaturization and individualized development of the agricultural machinery makes some of the disabled can participate in agricultural activities by operating the machine.A gait assistance device can assist the lower-extremity disabled to finish work more successfully and efficiently.The gait assistance device can provide assistance during gait according to the gait trend of the users.Therefore,we developed a gait-phase recognition method for designing a gait assistance device.Surface electromyography signal(sEMG)have a close relationship with muscle activities and motor functions of people.sEMG signals have the ability to reflect the movement intention of people and the state of muscle.It can be used to gait phase recognition.The different gait sub-phase can be recognized due to their different characters of EMG.The original sEMG signals were pretreated using filter and denoising technology Then,the gait phases of one gait cycle were recognized by using suitable feature extraction and pattern recognition methods.The main works of this study include:(1)The pretreatment methods of sEMG were analyzing for their high signal to noise ratio,nonstationary and feeble characteristic.The original sEMG signals were filtered using the wavelet denoising technology and a band-pass filtered at 10-350 Hz.(2)The feature extraction methods of sEMG were analyzing for the gait phases recognition of the motion defected person.The time domain,frequency domain,time-frequency domain and nonlinear dynamics features are extracted after the signals are pretreated.The feature sets were researched for improving the recognition accuracy.The feature sets are selected for gait phases recognition.(3)Constructing the five-LIBSVM classifier for recognizing the different gait phases of disabled people.The extracted features are used to recognize the five gait phases in LIBSVM.The analyze results indicated that the recognition results of Mean absolute value(MAV)&Zero crossing(ZC)are the best.(4)Statistical methods are used to analysis recognition results of baseline and post-adaptation in the three kinds of resistance for analyzing the generalization of the proposed pattern recognition methods.The recognition results of the three kinds of resistance are also analyzed by statistical tool.The three kinds resistance load are Abrupt,Gradual and Noise.The statistical analysis results indicated that the gait sub-phase detective difference between the baseline and post-adaptation was not significant except the results of pre-swing in gradual(p=0.031).The statistical analysis results indicate that the gait sub-phase detective difference among the three resistance load conditions was not significant.This research focus on the gait phases recognition of lower limb disable person.The research includes pretreating using filter and wavelet de-noising technology,feature sets research and gait phases recognition.The suitable feature sets for lower limb disable person are selected through the sEMG signals analyzing.The five gait phases are recognized by LIBSVM.Furthermore,the statistical analysis methods are applied to the gait recognition results.Achievements in this study will play important role in evaluating the lower limb's function,developing new intelligent assistance device lower limb.