Design Of Surface EMG Signal Acquisition Armband And Research On Gesture Recognition

Human gestures can express a lot of meanings.After the computer recognizes the control information in gestures,the corresponding commands can be executed to achieve the purpose of natural human-computer interaction.SEMG is a kind of bioelectrical signal produced by muscle contraction,which contains information such as the subjective intention of the movement sender.Compared with traditional gesture recognition,gesture recognition based on SEMG signal has many advantages,such as convenient acquisition and non-environment sensitive.At the same time,the acquisition,processing and transmission of SEMG signal are challenged due to its own characteristics.In this paper,SEMG signal processing and gesture recognition methods for wearable applications are studied.Firstly,for a large number of noisy signals collected in daily wear applications,signal pretreatment is carried out based on methods such as least square method,IIR data filtering and six-layer wavelet,etc.,which can effectively deal with the problems such as baseline drift,power frequency interference and jitter in the original signal and achieve high-quality signal envelope extraction.Secondly,a gesture recognition method based on discrete value is proposed for the real-time and low power consumption of wearable gesture recognition.Finally,in view of the shortcomings of the existing SEMG acquisition armband on the market in the openness and convenience of data,an experimental SEMG acquisition armband was developed for the evaluation and testing of the above methods.The main work is as follows:1.In the signal preprocessing part,the least square method is adopted to deal with the baseline drift;Chebyshev Type I IIR high-pass filter with cut-off frequency of 20 Hz was used to process the signal motion artifacts.A second-order IIR digital notch device with a notch frequency of 50 Hz was used to deal with power frequency interference.The six-layer wavelet decomposition and reconstruction method is used to smooth the envelope signal.Simulation results show that the signal preprocessing method can greatly improve the accuracy of gesture recognition.2.In the gesture recognition part,this paper proposes a gesture recognition method based on discrete values.For the SEMG signals of the typical position of the arm collected by multiple channels,the original EMG signals were preprocessed and the active segments were segmented.Then the data of each channel is discretized according to the method proposed in this paper.Finally,the discrete values obtained from all channels are matched with the patterns in the training for gesture recognition.The experimental results of gesture threshold and channel number show that the proposed algorithm can effectively recognize the predetermined six gestures.3.In terms of experimental system development,this paper made a myoelectric armband according to the demand.The whole armband system is introduced in detail from three aspects: armband structure,hardware system and upper computer.Through the EMG armband experimental device,the acquisition position of the arm can be adjusted and the acquisition of SEMG signals can be carried out without restriction,which can fully support the evaluation and testing of various methods and algorithms proposed in this paper.4.Based on the SEMG armband experimental device,this paper carried out SEMG acquisition experiments,compared three electrode placement schemes,and verified the effectiveness of the proposed armband structure by feature acquisition of the three placement schemes.Experimental results show that the gesture recognition algorithm based on discrete values can achieve high recognition accuracy and speed,and the armband structure designed in this paper is conducive to SEMG signal feature extraction.