Research On The Key Technology Of Wireless High-Throughput EEG Signal Acquisition And Its Application In Brain-Computer Interface

Brain-computer interface(BCI)system can directly exchange and communicate information between the brain’s neural activity and external computers and devices.Its research and application are developing rapidly,which has a profound impact on many fields such as neuroscience,medicine,computer science and engineering.This paper is aimed at exploring bioelectrical signal detection technology for health monitoring and BCI application,breaking the technical bottlenecks of the system in miniaturization,portability,multi-collection channels,high throughput data communication.A miniature wearable intelligent EEG acquisition &processing system was developed to improve the effectiveness of the system in practical applications.The specific research content of this paper includes:1.The portability,wearability and high-throughput detection of BCI devices has been a major obstacle in their application.A 64-channel high-speed wireless EEG acquisition system based on FPGA was developed to solve the technical bottlenecks in volume,power consumption,bandwidth and sampling rate.The developed EEG system can support both intrusive and non-intrusive acquisition of all-frequency EEG signals,due to the difficulties of low noise amplification of small signals,high-speed sampling and high-speed wireless transmission was overcomed.Based on the complex programming technology of FPGA and ARM dual-core processor,the wireless high-throughput stable transmission of sampling data was realized through high-speed USB interface.The system achieved a maximum sampling rate of 30 k S/s per channel and a maximum data transmission rate of 58 Mbps.In this paper,the system can support invasive and non-invasive acquisition of full-frequency domain EEG signals,and has the advantages of excellent performance indicators,small size,and high integration.2.Based on the developed high-speed wireless EEG acquisition system,the EEG acquisition of primary motor cortex in macaque monkeys was performed and compared with the commercial Omniplex EEG acquisition system.To solve the problem that the artificial extraction of neuronal Spikes is easily affected by subjective factors,and the low accuracy of automatic threshold extraction,the root-mean-square threshold detection algorithm was optimized to realize the threshold detection of Spikes signals by setting the root-mean-square threshold of signals within the sequence duration,and the detection accuracy of Spikes sorting of macaque neurons was improved.A classification algorithm based on quasi-recurrent neural network(QRNN)was proposed to classify the grasping task of macaque monkeys,using the advantages of network for spatial recognition and sequence modeling,due to achieving a classification accuracy of 98.03%.In addition,electrical stimulation of the primary motor cortex of macaque monkeys was also conducted,and the effect of neuronal activity in this brain region on the behavior of was investigated.3.The developed wireless high-quantitive EEG acquisition system was applied to steadystate visual evoked potential(SSVEP)classification of EEG signals.To study the feasibility of the developed EEG system in non-invasive EEG signal acquisition,the collected SSVEP signals were classified by the anonical correlation analysis(CCA)algorithm,and compared with the commercial Grael EEG system.According to the analysis of classification accuracy and standard deviation,the self-developed system has higher detection performance.Further,because the classification accuracy was affected by the differences between different subjects caused by the fixed window duration in the task correlation component analysis algorithm,an adaptive window duration method based on covariance analysis was proposed,which combines CCA and TRCA algorithms.The window duration of TRCA algorithm was dynamically adjusted according to the signal characteristics of the subjects,and the classification accuracy of 95.12± 3.92% and ITR of 82.76±4.36 bits/min were achieved.4.To solve the problems of inconvenient wearing,unfriendly user experience and easy to be disturbed,a high-performance wearable wireless ear canal EEG(Ear-EEG)device were developed.The 16-channel wearable Ear-EEG acquisition system was developed based on high-density ear electrode and high-integrated electrophysiological signal interface chip,a high-density array ear electrode composed of a gold-plated spring thimble electrode and a custom-made silicon gel ear mold was designed,and the high signal-to-noise ratio(SNR)acquisition was realized.The developed Ear-EEG acquisition terminal has the advantages of high performance,low power consumption and small size.It can support a sampling rate of 1.5k S/s per channel,provides basic support for the development of Ear-EEG based BCI system and its application in non-laboratory environment.5.The classification of SSVEP signals high-performance in ear canal was studied by using the self-developed EEG acquisition system,and the classification results were compared with those of commercial scalp EEG system.The application performance of the ear canal EEG acquisition system was verified in this paper.To solve the problem of weak SSVEP signals in ear canal,a convolutional neural network model based on parallel multi-scale filter banks was proposed.By adjusting the multi-scale filter banks and training the deep learning model,SSVEP signals were recognized.Finally,the classification accuracy of SSVEP signal in ear canal reached 88.06±3.58% and the ITR reached 61.73±7.34 bits/min.