Analysis of brain signal data like Electroencephalography (EEG) plays an important role in efficient diagnosis of neurological disorders and treatment. EEG records electrical activity of the brain and contains huge vo...
详细信息
ISBN:
(纸本)9783030908881;9783030908874
Analysis of brain signal data like Electroencephalography (EEG) plays an important role in efficient diagnosis of neurological disorders and treatment. EEG records electrical activity of the brain and contains huge volume of multi-channel time-series data that are visually analyzed by neurologists to identify abnormalities within the brain, which is time-consuming, error-prone, and subject to fatigue. Therefore, an automatic data mining system is always in need to detect abnormality from those large volume of data. To meet the requirements, in this study, a time-frequency spectrogram image-based classification framework is developed using texture feature extractor and machine learning (ML) based classifiers. At first, signals are filtered to remove noises and artifacts and normalized. Then signals are segmented into small chunks and spectrogramimages are generated from those segments using short-time Fourier transform. After that, histogram based textural features are extracted and significant features are selected using principal component analysis. Finally, those features are fed into three ML based classifiers for categorizing the signals into different groups. The proposed system is tested on EEG brain signal data and have obtained promising results in identifying different abnormality groups, which indicates that the proposed system can be used for mining large volume of brain signal data.
暂无评论