Research On Features Extraction Based On Speech Imaginary Recognition

The brain-computer interface(BCI)is a direct communication channel between the human brain and a computer(or other control device).At present,there are two main types of BCI system.One is a BCI system based on induced EEG,such as P300-based BCI,SSVEP-based(steady-state visual evoked potential)BCI.This type of BCI recognition rate is high,and the research is relatively mature.The BCI does not require user training,but requires specific equipment for users with stimulation.Under the stimulation of external equipment users are easy to feel fatigue,and the adaptability is more limited so this type of BCI is more suitable for communication.The other is a BCI system based on spontaneous,like ERD/ERS-based(event-related desynchronization / synchronization potential)brain-computer interface.The braincomputer interface does not need external stimulus,only need the user to think and need to go through a long time training.The classification accuracy need to be improved,and the brain-computer interface is more suitable for control.Based on the background,a new brain-computer interface model-the brain-computer interface based on language imagination-comes into the researcher's vision.Language imagination is the use of spontaneous EEG,do not need external stimuli,and do not need long training.Its accuracy is able to achieve the required accuracy and the brain-computer interface has extensive research significance.This thesis focused on the study of feature extraction based on speech imaginary.In the process of studying speech imaginary,a speech imaginary paradigm of phonetic vowels was proposed,and the subjects experimentally verified the paradigm.On this basis,a single imagination model and a mixed imagination model were proposed,and we analyzed the effects.In the study of the EEG signal recognition,the EEG signal was band-pass filtered,and the feature extraction adopts the method of common spatial pattern(CSP)and band power.The recognition model of BCI is established by linear discriminant analysis.This thesis achieved the following research accomplishments:1.A new model of speech imaginary was proposed,and the phonetic alphabet of Chinese characters was studied.Eight subjects were used to collect EEG signals from off-line experiments.2.A brain-computer interface model for language imagination was established,and EEG signals were processed to select valid features and classified them.The method of recognizing EEG signals based on phonetic alphabet was obtained.3.The single imagination model and the mixed imagination model in the process of speech imaginary were put forward.The experiment was carried out in two cases,and the experimental results were verified by the BCI model of speech imaginary.The experimental results showed that the mixed imagination model can improve the subject 's attention more than the single imagination model,and the classification effect is better and more suitable for the BCI system.