Deep learning mixed model-based steady state visual evoked potential classification method

Abstract

The invention discloses a deep learning mixed model-based steady state visual evoked potential classification method. The method comprises the steps of 1, adopting an LCD display as a stimulation source, determining a flicker frequency, selecting an electrode channel for electroencephalogram collection, carrying out an experiment for multiple different testees, and performing collection to obtain a steady state visual electroencephalogram signal database; 2, based on short-time-sequence electroencephalogram signals in the database, training and determining parameters of a convolutional neural network model, and finishing automatic extraction of features of the electroencephalogram signals; and 3, adopting an output of a convolutional deep learning network as an input of a Boltzmann machine network, performing fine adjustment on parameters of a classification network model for the different testees, and determining parameters of a Boltzmann machine network model. According to the method, the extraction of the generalization features of the electroencephalogram signals can be well realized; the influence of electroencephalogram signal distortion on signal classification is reduced; and the short-time-length electroencephalogram signals can be utilized to well finish the signal classification.

Description

technical field [0001] The present invention relates to the technical field of brain-computer interface, in particular to a steady-state visual evoked potential classification method based on a deep learning hybrid model. Background technique [0002] The brain-computer interface is a system that translates the EEG signals of the human brain into external control signals, opening up a new channel for the information exchange between the human brain and the outside world. As a new type of human-computer interaction, the brain-computer interface is gradually becoming a hot topic in brain science research, and has great application prospects in the fields of rehabilitation engineering, high-risk operations, and psychological cognition. In recent years, with the development of brain science, signal processing technology, computer science, etc., brain-computer interface technology is developing rapidly. [0003] According to whether the generated process requires external stimul...

AI Technical Summary

Problems solved by technology

[0005] Because the EEG signals collected by the electrodes on the surface of the scalp are extremely weak, usually more than ten millivolts, and the background noise such as myoelectricity, spontaneous EEG, and power frequency interference are much larger than the EEG signals, this will cause a steady state induced brain disturbance. Classification of electricity presents difficulties

At present, the difficulty in researching algorithms for the classification of steady-state visually evoked EEG signals lies in: first, steady-state visually evoked potentials will be distorted between different experimental subjects, and the classification effect is different for different subjects; second, the main considerations The characteristics of the signal frequency domain, the feature selection is relatively simple; third, the balance between the signal time length used for classification and the classification accuracy rate, often the longer the signal time used for classification, the higher the accuracy rate will be

Images