Effects of Audio and Visual Stimulation on EEG in Sleep Assessment and Brain Computer Interface

• Main Authors: Aaron Raymond Ang See, 施金波
• Other Authors: Shih-Chung Chen
• Format: Others
• Language: en_US
• Published: 106
• Online Access: http://ndltd.ncl.edu.tw/handle/h838av

博士 === 南台科技大學 === 電機工程系 === 102 === Audio stimulation and video stimulation (AVS) are methods usually used in neuro-feedback research with a wide range of applications. There are various applications of AVS but we focused our study to the use of AVS on two main fields of applications namely sleep quality assessment and brain computer assessment (BCI). In this dissertation, we will assess the effects of audio-visual stimulation (AVS) on EEG under various conditions. First, we validated the coherence of the prefrontal EEG electrode placement for sleep EEG measurement. To achieve these goals, we ascertained its validity through the use of statistical methods and manual verification. Second, the pre-frontal EEG electrode placement is utilized for sleep quality assessment. The current study would perform six tasks which include an AVS program and acquire signals using the prefrontal EEG and evaluate the short-term changes on the EEG signals of subjects who went through a good sleep or poor sleep condition. Third, the effect of melancholic music listening on the prefrontal EEG would also be discussed. Fourth, visual stimulation using steady state evoked potential (SSVEP) was used for our BCI system that was applied for both communication and entertainment purposes. In this study, we will focus on the evaluation and improvements of existing BCI systems. Lastly, speech signal analysis, in the form of cepstrum, will be tested in the feature extraction and classification of the BCI system for future use. Results from the different studies of ours demonstrated the following outcomes: First, a strong coherence at different frequency harmonic bands with an average correlation of 0.947±0.01 between the traditional sleep EEG electrode placement and the alternative electrode placement. Further, manual sleep staging also manifested strong congruence between the 2 different electrode placements. Second, the theta (θ) waves obtained from the pre-frontal EEG manifested a significant difference between good and poor sleepers specifically during the audio stimulation task displaying a 25% difference in mean theta amplitude. Third, melancholic/sad music did not show any significant changes in the pre-frontal EEG asymmetry of normal subjects. Fourth, the BCI spelling system tested by 7 users exhibited an average accuracy of 92.80% and a peak information transfer rate (ITR) of 21 bits/min. Fifth, the BCI remote-control entertainment system, evaluated by 12 subjects, had an average accuracy of 89.51% and 92.31% for the training and simulated path test respectively. The highest ITR obtained by any user in the 2 tests was 105 bits/min. Sixth, power cepstrum analysis demonstrated a promising method by exhibiting a 92.58% accuracy using only 7 cepstrum coefficients. Our study has successfully explored the use of AVS for sleep diagnostics and sleep quality assessment using pre-frontal (forehead) EEG. Nonetheless, an innovation on a convenient and practical EEG device is still a key factor for more EEG studies to be easily implemented. We have also objectively assessed our BCI systems and deliberated the use of cepstrum analysis and GMM for our BCI applications. Subsequently, there is a need to develop an efficient stimulator, and standalone system for BCI applications. This dissertation has provided a broad understanding on how AVS affects EEG in both sleep assessment and BCI applications.