Mini-grid Sensor System For Surface Electromyography Acquisition And Neural Information Decoding

Surface electromyography(s EMG)is now a widely used biosignal for the human-machine interface,especially for prosthetic manipulation.The dexterous prosthesis is useful to reconstruct the limb function of the amputee.However,the mainstream control strategy of the dexterous prosthesis,pattern recognition,could suffer from performance degradation in practical applications due to factors such as electrode shift and muscle fatigue.In recent years,the motor unit decoding algorithm based on high-density surface EMG signals has become hotspot because it could obtain the nerve drive information of human motion.High-density EMG acquisition systems are usually complex,bulky,and expensive.It is challenging to integrate HD-EMG systems into the prosthesis.On the other hand,the conventional multi-channel EMG acquisition system could barely support the surface EMG decomposition algorithms due to the insufficient spatial resolution.To address the problems,this paper carried out the following works:Aiming at improving the spatial resolution of portable s EMG acquisition devices,we designed the mini-grid s EMG sensors.By arranging multiple channels on a single sensor,the local spatial resolution of the system is improved.Furthermore,a portable EMG acquisition module is designed to constitute the hardware of the EMG acquisition system.Each module could be connected to four sensors synchronously,which means it could provide16 channels of EMG signals.The system performance tests verified that the system could provide a high quality of EMG signals as existing commercial systems.Based on the C# programming language,the EMG signal acquisition platform software was developed.The software provides EMG signal reception and storage,real-time signal display,experiment integration,and signal decoding functions.Meanwhile,we provide a software development kit(SDK)for software customization.An s EMG decomposition algorithm based on the mini-grid acquisition system is implemented.The s EMG signals are split and clustered to generate MUAP templates.The discharge timings of motor units are finally determined based on cross-correlation.The algorithm was tested with simulated signals and real EMG signals.However,the accuracy of the algorithm remains to be improved.Based on the mini-grid EMG acquisition system,we test the performance of surface EMG decomposition on motionless hand gesture recognition and compared results with the pattern recognition method.The results revealed that hand gesture recognition with MUAPt could achieve an equivalent accuracy to the pattern recognition method.The mini-grid surface EMG acquisition system proposed in this paper has a comparable signal quality to that of commercial systems but is more portable and economical.It is easy to integrate with prosthetics.Subsequent gesture recognition experiment provides a preliminary proof of the possibility of decoding neural information from multi-channel systems.