| The brain-computer interface(BCI)is an information transmission channel which is created directly between the human brain and external devices without relying on the peripheral nervous system and muscle tissue.It has tremendous practical value in various aspects especially in the fields of medical rehabilitation and operation in special environments.The BCI system is mainly composed of four parts: signal acquisition,feature extraction,pattern recognition and control signal output.The essential technology is a conversion algorithm that decodes the input EEG signals into control signals.The " Imitating Reading " brain computer interface(IR-BCI)was first proposed by the research team of South-Central University for Nationalities.Further more IR-BCI has achieved some good results in terms of information transmission rate and recognition accuracy.A domestic team has developed an online IR-BCI system for real-time character output.However,the EEG signal processing algorithm in the system depends on the PC which makes the IR-BCI system greatly affected in portability and practicality.Therefore,the current application results in the laboratory stage.With the development of programmable logic device technology,FPGA has been greatly optimized in data processing performance and operation speed,making it practical for FPGA to handle most of the data flow control and algorithm tasks in the BCI system.This paper designs a FPGA-based IR-BCI hardware system which transplants the data preprocessing and core algorithms in the existing IR-BCI system to the lower computer.It hasgreatly improves the portability of the system and promotions for the practicality of IR-BCI technology application.The principal work of this article includes EEG signal acquisition,data preprocessing,and core algorithm transplantation based on FPGA platform.The signal acquisition consists of brain electrodes and A / D modules,in which the A / D module is controlled by the FPGA controls the chip ADS1298 to complete data transmission.Data pre-processing contains band-pass filtering,feature extraction,data down-sampling and normalization,in which the band-pass filtering is finished with IIR band-pass filter;the time-domain method is used for special diagnosis;and the normalization takes the maximum and minimum normalization method.The core algorithm transplantation is made up of the LS-SVM.Based on the existing methods,the real-time calculation function of the neural network weight coefficients in thealgorithm on the FPGA is innovatively designed to make it better applied to real-time BCI system.Finally,in order to test the working status of the system,the functional results of each module of the system are tested with the PC end processing results as a reference.The test results show that each module can work normally.The accuracy of the sample classification of this system is almost the same as that of the system based on the PC platform,which meets the design requirements and verifies the feasibility of replacing the PC with FPGA to complete the data processing. |
