FPGA-based rapid classification method, implementation method and device for EEG signals

Abstract

The invention discloses a fast classification method of EEG signals based on FPGA, a realization method and a device thereof. The hard logic of a CNN network structure model suitable for the classification of EEG signals is constructed on FPGA, and the convolution operation is converted into matrix multiplication. The IP cores of each layer in the CNN network structure model are established, and the IP cores of each layer in the CNN network structure model are connected by using the synchronous data flow method, and AXI4-Streaming register chip is inserted between the adjacent IP cores; the EEG training data is received, the floating-point data is converted to a fixed-point number with a preset number of bits, training the CNN network structure model, adjusting the CNN network structure model weight value until the highest classification accuracy model is obtained, and the trained model parameters are stored in DDR memory, so as to obtain FPGA which can realize the fast classificationof EEG signals, and the fast classification of EEG signals is carried out by using the CNN network structure model.

Description

technical field [0001] The technical field to which the present disclosure belongs is brain-computer interface (BCI), and relates to an FPGA-based fast classification method, implementation method and device for EEG signals. Background technique [0002] The statements in this section merely provide background information related to the present disclosure and do not necessarily constitute prior art. [0003] Brain-computer interface technology involves the fields of computer communication and control technology, neurology, biomedical engineering, and rehabilitation medicine. It interprets user intentions by analyzing and processing EEG signals. At present, EEG signals generated by different stimuli can be collected through special equipment. To directly control peripheral devices through the brain, it is necessary to accurately identify and classify the EEG signals collected by the equipment. [0004] At present, the acquisition mode of EEG signals is mostly realized throug...

AI Technical Summary

Problems solved by technology

[0007] Aiming at the deficiencies in the prior art and solving the problem of how to balance flexibility and reconfigurability for different network structures in the process of EEG signal classification, one or more embodiments of the present disclosure provide an FPGA-based EEG A fast classification method, implementation method, and device for electrical signals, realizing fast classification of EEG signals with deep convolutional neural networks taking into account scalability, flexibility, and reconfigurability

Images