Motion Intention Recognition Based On EEG2Image Denoising Convolutional Neural Network

Applications in the internet of things(IoT)are promising,as a large number of smart environments can be connected to BCI.BCI is a real-time communication system that connects the human brain and external devices,directly converts the information generated from the brain into commands for driving external devices,and replaces the human body or speech organs to communicate with the outside world.In other words,the BCI system can replace the peripheral nerve and muscle tissue to realize the communication between people and external environments.The main technical problems of the BCI system include the feature extraction of EEG signals,the feature representation of EEG signals,and the classification of EEG signals for specific intent activities.The Brain-Computer Interface for Motion Imagination(MI-EEG)is one of the typical applications.By using the motion imaging BCI system,the controller can realize the expected action without manual operation,such as mind control of the wheelchair,EEG control of the exoskeleton,mind input and help patients with physical disabilities to restore arm movement.This kind of mixed intelligence is used as an auxiliary rehabilitation training for the poor physical condition or the disabled,as a carrier of perception and intervention in animal behaviour and control.In short,BCI system-related research collects and analyzes brain-related signals,detects and interprets brain states,such as brain intent,context translation,emotional state,etc.,thereby providing the possibility to discover the relationship between brain activity and human behaviour.The critical problem of the non-invasive BCI system is to accurately identify the person's intention.However,the low signal-to-noise ratio,low classification accuracy,and poor generalization ability of electroencephalogram(EEG)and other EEG signals limit the application of non-invasive brain-computer interfaces in practical applications.In recent years,the analysis of EEG signals through non-invasive BCI systems has attracted widespread attention in the field of pattern recognition.In this paper,the relevant data sets of motion imagination experimental paradigm were studied,and an analysis framework for motion intention recognition based on EEG2 Image denoising convolutional neural network was proposed:This paper proposes a denoising convolutional network model based on EEG images,which can learn the colour and spatial changes of image objects converted from EEG signals.First,the original EEG signal is decomposed into different frequency bands and sliced using a time sliding window to reduce noise in other frequency bands.Then,this article further converts the sliced EEG signals into RGB images.Secondly,this paper uses the denoising convolutional network structure to learn the colour space and spatial changes of image objects,which means that it is no longer necessary to accurately crop new training images.The colour space transformation and the space transformation network realize the transformation of the colour space and the expansion of the colour area on this basis.Therefore,the training process is greatly simplified,and you only need to focus on adjusting the hyperparameters in the network to obtain optimal performance.The results show that the method proposed in this paper has achieved the best results currently on this type of data.