Feature Extraction And Classification Of Eeg Based Imagery Left-right Hands Movement

In recent years, as electronic information, computer science, the rapid development of EEG signal processing theory and techniques needed to become more perfect, and therefore brain-computer interface (Brain.Computer Interface, BCI) research started to become hot. Brain-computer interface research in different ways, based on imagined movement of brain-computer interface system for its simple, effective, non-invasive, and cause for concern. BCI systems studied in this paper is to imagine the movement based on classification of EEG research, the ultimate aim is to use the spontaneous characteristics of the human brain to distinguish the specific task of the brain thought to form control commands to achieve between computer and human exchange of information and hope to have a physical action on the rehabilitation of persons with disabilities health disorders that EEG-based brain computer interface is simple, you can not rely on external stimulation device, but also easy access to downtown electrical signal source, which has good prospect. However, the spontaneous EEG signals are very weak, noisy, and non-stationary signal, so efficient methods of EEG analysis is the core of BCI system. In summary, based on the work of predecessors in the past, the paper EEG acquisition and processing, research on the following aspects:Stimulator using a new rapid access to brain signals:The original brain signal acquisition process: an independent mental tasks corresponding to an EEG acquisition process, repeat mental tasks can get all the EEG samples. This disadvantage is one of the largest distributed EEG acquisition cycle, the time required is too long (one signal acquisition experiments often have to spend a whole day), easy to make the experimenter tired of thinking, thus affecting the normal completion of the task of image movement. This EEG acquisition process: a corresponding number of independent thinking, EEG acquisition process work. It is accomplished through the stimulation to overcome these shortcomings.Signal preprocessing:EEG with high noise, spontaneous EEG signal amplitude range of more than ordinary weak brain, it is necessary to filter the original signal preprocessing, as far as possible to eliminate noise interference. This will remove the DC-EEG were used after the median filtering, IIR filtering to remove the eyes were electric, muscle and other sudden noise, and the signal band-pass filter.Feature extraction and Classification:EEG is a typical non-stationary signals, from the time domain or frequency domain EEG feature extraction, in accuracy and can not meet the requirements. In this paper, short-time Fourier transform (STFT) time-frequency feature extraction of EEG, and through the cycle map estimation is estimated by the short time Fourier transform power spectrum of the signal after each paragraph, and the last movement combines the unique imagination ERD / ERS (literature 1) as a feature vector for final classification and identification tasks.