Analysis Method Of EEG Signal In Alcoholics Based On Normal Time-frequency Transform

Electroencephalograph(EEG)is one of the main techniques to study brain function by recording electrical signals on the scalp of the brain.It has been widely used in diagnosis of mental diseases and in the field of brain-computer interfaces.EEG signal belongs to unstable random signal with uncertainty and complexity.The processing and analysis of EEG signals has become a research hotspot in recent years.Time-frequency domain analysis method is one of the main analysis methods for analyzing EEG signals.However,the time-frequency information obtained by traditional time-frequency transform methods contains pseudo frequency which affects the accuracy of the results.In addition,some EEG signal analysis methods are sensitive to parameters,and introduce many features which result in complexity in data processing.In order to solve this problem,an EEG signal analysis method is proposed based on normal time-frequency transform.EEG database sampled from alcoholics are used to test the proposed method.The research topic is of theoretical and practical importance.The main contributions of this paper include:(1)The acquisition process of EEG signals of alcoholics is described,and the commonly used EEG signal analysis methods are briefly introduced.In order to remove the interference by physiological electrical field and equipment during the data acquisition process,the EEG data are preprocessed,including single event extraction,artifact removal,bandpass filtering and average reference electrodes.The processed EEG data will be used as training samples and testing samples for subsequent analysis.(2)A novel time-frequency analysis method of EEG signal is then proposed.It is based on the normal morlet transform in the normal time-frequency transform theory,which is capable of dealing with the problem of pseudo frequency occurring in the traditional time-frequency transform method.Therefore,more accurate information about the frequency,amplitude and phase of EEG signal can be obtained.In order to solve the problem of edge effect in the normal time-frequency transform,the edge artifacts in the time-frequency domain of the EEG signal are removed.The amplitude in the time-frequency domain of the EEG signal is then statistically analyzed.It is found that the the EEG signal of the alcoholic and normal people mainly differs in the ?,? and ? wavebands,and the alcoholics have a lower amplitude value.It shows that the normal time-frequency transform method is effective in the analysis of EEG signals.(3)In order to solve the problem that most traditional EEG signal classification methods are sensitive to parameters,and introduce many features which result in complexity in data processing,an EEG signal classification method is proposed based on features extracted by normal time-frequency transform.Firstly,the average amplitude of each waveband is obtained by use of normal time-frequency transform,which is then applied as input to a KNN classifier to classify the EEG signals of the alcoholics group and normal groups.Next,the proposed method is tested by 10-fold cross-validation method,and the influence of window width parameters on time-frequency diagram and the classification accuracy is analyzed.The experimental results show that different window widths have no significant effect on the final classification accuracy.The classification accuracy rate varies between 98.4%-99.2% with varying window width parameters,indicating the effectiveness of the proposed method.(4)The proposed method is compared with other EEG analysis methods such as traditional morlet transformation,short-time Fourier transform and wavelet transform.The results show that the proposed method has better robustness to parameters,less classification features and high classification accuracy.Result of the present work provides a new idea for the processing and analysis of EEG signals.