Research On Surface EMG Signal Detection And Novel Processing Method With Flexible Electrodes

With the rapid development of intelligent computing technology,the exploration and research of novel human-machine interaction technology emerge in an endless stream.Gesture recognition technology is a hot topic in the field of Human Computer Interaction in recent years.Surface Electromyography(SEMG)is an important biological signal generated from muscle activity.The specific joint movements of the limbs are controlled by their corresponding muscle groups.The electromyographic signals generated during exercise can reflect the body posture and muscle activation patterns.It is widely used in the field of human-computer interaction,sports medicine,rehabilitation engineering and so on.Because the SEMG signal is a weak physiological signal,it is easy to couple external noise,which causes great interference in signal acquisition and processing.Currently,SEMG signal detection usually use wet electrodes or metal electrodes,which is not comfortable enough for long time wearing.So we need to use new electrode materials instead of traditional electrodes to collect SEMG signals.The flexible fabric electrode can be well fitted to the human skin topology,and can effectively suppress the motion artifacts.So it is suitable for wearable physiological signal acquisition system.To sum up,this thesis will study the acquisition and processing method of SEMG signal based on flexible electrodes for long-term monitoring.The main work and innovation of this article include:(1)A wearable surface SEMG acquisition and processing hardware platform based on the flexible fabric electrode is designed.The suitable flexible electrode material is selected to extract the SEMG signal;And then through the acquisition front-end,the SEMG signal is amplified and filtered;The micro controller collects the SEMG signal after the adjustment,and carries out the wireless transmission through the data transmission module.According to the wearable features of wearable equipment,the hardware circuit with high integration and strong stability is designed.(2)The hardware drive part realizes the acquisition of SEMG signal,the signal preprocessing,the data format conversion and the wireless real-time transmission of the signal.A real-time surface electromyography acquisition and analysis software is designed,include data reception,data analysis,signal processing,feature extraction,feature training,pattern recognition,real-time and historical data display,data storage and other functions.(3)The dynamic SEMG signal is processed and analyzed.It include the design and contrast of signal processing,data segmentation and feature extraction of SEMG signal,and proposes a gesture recognition scheme based on SEMG.The appropriate filter for the common noises in SEMG signals is designed.Data segmentation is improved for moving average method based on valve domain,which reduces the incidence of segmental errors caused by short time of rest signal in continuous gesture signals.Feature extraction involves the time domain characteristics,frequency domain characteristics and timefrequency characteristics of surface electromyography.Also the gesture recognition effect is verified.(4)The best selection of optimal electrode channel and electrode position is analyzed.Because the human upper limb muscles have really complex structure,the distribution of many fine muscle finger and hand movements are similar.It results that the SEMG signal collected has little difference.Placing an effective part of the forearm pole can greatly reduce the computing and data transmission amount,reduce the computational complexity of the algorithm of gesture recognition,so an experimental study with cross validation and comparison on the effect of electrodes whith different numbers and positions on the classification results.Combined with the MRMR-FCO(minimal Redundancy-Maximal Relevance method with F-test Correlation Out)method,the features of surface electromyography and inertia signal are sorted and feature compression is done.Combined with accuracy and real-time,the best feature extraction scheme is designed.