Tomonari Yamaguchi

Wavelet analysis of EEG signals during motor imagery

ERS and ERD (event-related synchronization and desynchronization) are observed in EEG (electroencephalogram) signals around such events as sensitive stimulus, motions, cognitive actions etc. Usually, ERS/ERD features of EEG are extracted as variances of band-passed signals of several trials. To make use of these features to recognize inputs for BCI (brain-computer interface), we applied discrete wavelet analysis to extraction of ERS/ERD features from a small number of EEG signals during motor imagery. We employed Daubechies, convolution and spline biorthogonal mothers for linear wavelet analysis, also Haar type structural function for morphological wavelet analysis. Then our extraction method was estimated by the pattern recognition based on AR model.

Feature Extraction of EEG Waves Based on Morphological Multiresolution Analysis

Electroencephalograph (EEG) recordings during right and left hand motor imagery can be used to move a cursor to a target on a computer screen. Such an EEG-based brain-computer interface (BCI) can provide a new communication channel to replace an impaired motor function. It can be used by e.g., handicap users with amyotrophic lateral sclerosis (ALS). In this paper, feature extraction based on morphological multiresolution analysis is introduced to discriminate the EEG signals recorded during left and right hand motor imagery, and oddball task. The mixture features in the brain wave signal of single channel is able to separate to various bands. The structural function to decide the filter characteristic is discussed using the experimental studies.

REDUNDANT MORPHOLOGICAL WAVELET AND LOCAL PATTERN SPECTRUM

Time-frequency analysis methods such as wavelet analysis are applied to investigate characteristic from non-stationary signals. In this study, we proposed redundant morphological wavelet analysis that was a kind of nonlinear discrete wavelet and redundant wavelet. This method analyzes a transition of shape information from signals in detail since this method keeps property of shift invariance though information of decomposition includes redundancy. Local pattern spectrum which corresponds to nonlinear short time Fourier transform is derived from this nonlinear wavelet. The characteristics of these methods were confirmed by applying to simulation data and actual data.

Design Method of Morphological Structural Function for Pattern Recognition of EEG Signals During Motor Imagery and Cognition

Electroencephalograph (EEG) recordings during right and left hand motor imagery can be used to move a cursor to a target on a computer screen (such system is called BCI). Recently, we have proposed the detection method of Error Potential in order to add the fail safe function to BCI system. In this paper, feature extraction method based on morphological multi-resolution analysis is introduced to extract features concerned with motor imagery and cognition simultaneously from the EEG signals. Morphological filter is composed of nonlinear operation between signal and structural function and this multi-resolution analysis can be constructed by repeating this filtering to signal while changing structural function. This method is a kind of discrete wavelet analysis with non-linear characteristics and is effective to extract specific shapes. The structural function which decides the filter characteristic is designed to obtain optimal separation based on mutual information algorithm or spectrum dividing algorithm.

85. Feature extraction of EEG waveforms by directed information analysis during mis-operation of the brain–computer interface system

The aim of this study was to investigate the time series and space analysis of ictal high-frequency oscillations (HFOs) near the seizure onset area in mesial temporal lobe epilepsy (mTLE). All patients with mTLE were planned to undergo the intracranial EEG recording for presurgical evaluation. Depth and subdural macro-electrodes was inserted and digital EEG monitoring was done to detect ictal EEG. Besides routine procedure, EEG at 1000 Hz sampling was recorded for about 1 day. By visual inspection, HFOs more than 200 Hz was found in 41 seizures of 10 patients. For 60 seconds immediately before seizures, signals of all channels were analyzed with MBFA program. Almost all HFOs in 33 seizures were detected from channels around hippocampus or parahippocampal gyrus in the resection side. HFOs periodically appeared, but spectrum was consecutively found from 100 to 300 Hz with some prominent peaks. In many cases, HFOs evolved in occurrence rate and spectral power in a discontinuous manner. In three cases, HFOs seemed to be independent or lost during seizure propagation. In a case, HFOs showed different pattern in occurrence rate and spectrum between SPSs and CPSs. HFOs seemed to be related to the epileptogenic zone, but not expressed epileptogenesis directly.

EEG signal analysis during miss operation in BCI system

In this paper, electroencephalograph (EEG) signals recorded during miss operation of brain computer interface (BCI) system was analyzed. As a result, a feature difference between miss operation and normal operation was found at the low frequency component and the event-related potential (ERP) such as P300. Therefore a possibility of the modification of miss operation was confirmed.