Research Of Hand Gesture Recognition Based On Movement Imagination Brain Computer Interface

Intelligent manufacturing industry is the main direction of the "Made in China 2025" plan proposed by the State Council.As an important part of intelligent manufacturing industry,hand gesture recognition plays an important role in promoting the development of manufacturing industry.In recent years,hand gesture recognition based on brain computer interface(BCI)has gradually come into researchers' perspective.BCI uses EEG signals to build a bridge between the brain and the external environment,and does not rely on the human nervous system and muscle system."Brain science and brain-like research" of China's "13th Five-Year Plan" in the country's major scientific and technological innovation,who's importance is selfevident.However,the traditional recognition algorithms in the BCI have the problems such as "BCI blindness"(recognition accuracy rate is less than 70%),"one person one model"(recognition models cannot be shared due to individual differences),and difficulty in convergence during the recognition model training process.In this thesis,aiming at these problems,the experimental paradigm and recognition algorithm of BCI are built two BCI models of motion and imagination,and carries out experimental research combined with hand gesture recognition.The main contents and innovations of this thesis are as follows:(1)Aiming at the motion pattern,this thesis the experimental paradigm of the motion brain computer interface is proposed,and the hardware platform of the brain computer interface is built to decode the motion instructions in the brain of the subjects in the state of motion.For the problems of "BCI blindness" and "one person,one model" in the BCI,deep learning is applied to the classification and recognition module of the BCI.Based on the Long Short-Term Memory(LSTM),by improving the loss function,the Regularizer Long Short-Term Memory(RLSTM)and classification recognition algorithm are proposed.The experimental results show the accuracy of hand gesture recognition based on the motion BCI is up to 95.89%,and all subjects share the same recognition model,which solves the problem of "one person one model" and the difficulty of convergence of model training.(2)Aiming at the imagination pattern,this thesis puts forward the imagination brain computer interface experiment paradigm,builds the imagination brain computer interface hardware platform,and realizes the decoding of the brain thinking of the subjects in the static state.For the problems of "one person one model" and the difficulty of model training convergence,combined with the time-domain characteristics of the event-related desynchronization / synchronization(ERD / ERS)signal of the imaginary brain-computer interface,it is improved on the basis of LSTM and proposed LSTM-DENSE classification recognition algorithm.The experimental results show the accuracy of the BCI based on imagination is 91.56%,and the model training is easy to converge,and all subjects share the same recognition model,which solves the problem of "one person one model" and the difficulty of convergence of model training.