A multi-task motor imagery EEG feature extraction and pattern recognition method for vehicle control

Abstract

The invention relates to a vehicle control-oriented multi-task motor imagery EEG feature extraction and pattern recognition method, which uses an EEG amplifier to collect the multi-task motor imagery EEG signals of a subject, and transmits them to a host computer using welch power spectrum and a pair of A common spatial pattern algorithm extracts the combination of frequency domain and spatial domain features of motor imagery EEG; constructs multiple GMM classifiers according to the category of the training set data, passes the original EEG signal through the GMM classifier, and compares the obtained probability density with the set The credible threshold is compared, and the artificial neural network is used to perform secondary classification on the samples below the credible threshold, and the final classification result is obtained and transmitted to the vehicle through the wireless serial port to realize the real-time movement of the vehicle; the present invention utilizes the welch power spectrum Using CSP to extract frequency domain and space domain features related to motor imagery, using GMM and artificial neural network two-level classifiers, effectively improves the real-time control of vehicles and the safety of vehicle driving, laying the foundation for the practical application of brain-controlled vehicles.

Description

technical field [0001] The invention relates to a vehicle control-oriented multi-task motor imagery EEG feature extraction and pattern recognition method, and the fields of signal processing and pattern recognition. Background technique [0002] Brain-computer interface (BCI) can directly control peripheral devices through the interaction of neurons in the human brain. This control method that does not rely on limbs and peripheral nerves brings hope for the disabled to realize autonomous activities. Motor imagery EEG signals have the advantage of being evoked autonomously, so they are widely used in brain-computer interfaces. [0003] Motor imagery EEG signals have low amplitude and are easily affected by other physiological electrical signals, so their effective components are often submerged in noise. In order to improve the signal-to-noise ratio of EEG signals and extract the features that can best represent the subjects' imaginary actions, many frequency domain and spat...

AI Technical Summary

Problems solved by technology

However, its classification accuracy is still difficult to meet the high safety requirements in the actual driving process.

The artificial neural network contains a large number of weights and bias parameters. A large number of fitting parameters not only improves the nonlinear classification ability of the algorithm, but also increases its computational complexity.

Images