Research On Human-machine Interaction Method Based On SEMG Signals And Its System Design

Myoelectric signals,as a physiological signal closely related to human movement,can be applied to human-computer interaction technology.Myoelectric interaction realizes direct communication between humans and computers or other devices by using myoelectric signals as input.For example,people can control some machines by using gestures or movements.In the fields of metaverse,medical health and industry,myoelectric interaction has broad application prospects and huge market potential.However,there are still some shortcomings in the current myoelectric interaction system,which hinder the launch and promotion of myoelectric interaction technology.Therefore,this paper will conduct research on myoelectric interaction technology,discuss the current problems of myoelectric interaction and carry out a function-integrated surface myoelectric interaction system by combining deep learning technology and embedded technology.The main work is as follows:Deep neural network as a powerful algorithm model can be used to decode interactive information in myoelectric signals.However,the complexity of the most deep neural networks makes it difficult to deploy such algorithms on microcontrollers.Although neural networks have been widely used in various general-purpose computing platforms and information systems,their application in micro-embedded platforms is still quite imperfect.Therefore,this paper proposes a neural network model deployment framework for micro-embedded systems,named,emb AI(embedded Artificial Intelligence).The emb AI includes optimization deployment tools for network models and neural network inference engines running on microcontrollers.The emb AI can help micro-embedded systems quickly achieve end-to-end intelligent applications.Compared with mainstream X-CUBE-AI and TFlite-micro frameworks,the emb AI framework has a wider range of applicable platforms and lower porting difficulties.As far as the myoelectric interaction application discussed in this paper is concerned,although there is a gap of 17.52%～45.65% between emb AI and other frameworks in inference time,storage occupancy can be relatively reduced by68.27%～78.63%.That is to say,emb AI can run on more resource-constrained hardware platforms.This paper aims to integrate all the functionalities of a surface myoelectric interaction system into a microcontroller,which undoubtedly increases the load of the embedded system.Traditional embedded software system design frameworks tend to have more drawbacks when facing more complicated tasks.Moreover,current surface myoelectric interaction technology is still imperfect,so iterative improvement in the future may be necessary.Thus,the system design must consider reusability and scalability.To address these issues,this paper proposes a design framework for micro-embedded systems,which includes both software design strategies and hardware design strategies.The software framework,derived from object-oriented design principles,is an improved microservices-based design pattern.By abstracting functional logic and parameter details respectively as microservices and proxies,the software framework can give the system a clear design logic and make it easy to maintain and upgrade.At the same time,the hardware adopts a modular design philosophy and introduces an adapter design.The adapter can provide more flexible underlying hardware support for the embedded system,making it more versatile and feature-rich.In conclusion,based on the aforementioned research results,a function-integrated smart myoelectric interaction device was designed and implemented in this paper.This device can directly interact with many existing electronic information systems through Bluetooth protocol,rather than just being a simple myoelectric signal data acquisition device.Compared with myoelectric interaction systems that separate "acquisition" and "interaction",function-integrated myoelectric interaction systems have three advantages:(1)avoiding the driver adaptation work for different platforms;(2)extending the working time of wearable devices;and(3)enhancing the security of user privacy data.Experiments have shown that once the continuous transmission of myoelectric data to the interaction object is no longer needed,the energy consumption generated by wireless communication can be significantly reduced,and the average working time of the wearable device can be extended by 33.25%.