禪定腦電波頻帶判讀系統設計

Abstract

摘要 本論文的主要目的有兩個：(1)要設計一個禪定腦電波頻帶判讀系統，此系統要能夠判讀受測者的腦電波頻帶類別，並將結果以頻帶類別時序圖與統計表來表示(2)利用設計好的禪定電波頻帶判讀系統來判讀禪定者與非禪定者的腦電波，觀察兩者間有何差異。 在禪定腦電波頻帶判讀系統的設計方面，主要包含四個部分：(1)首先要蒐集受測者的腦電波資料，此步驟需要用腦電波錄製機來完成，本論文分兩組受測者來錄製，一組是禪定組，另一組是非禪定組(2)接著要萃取腦電波的頻帶特徵，本論文是以離散小波分解方式來得到腦電波的頻帶特徵(3)然後要利用萃取得到的小波特徵的空間分割來決定頻帶類別，並以此方法來找出適應性網路模糊推理系統(Adaptive Network-based Fuzzy Inference System ,ANFIS)的訓練資料的頻帶類別(4)最後要設計與訓練ANFIS，並驗證trained ANFIS的頻帶判讀性能，確定trained ANFIS具備頻帶類別判讀能力。經由以上四個步驟，整個禪定腦電波頻帶判讀系統設計才算完成。 在禪定腦電波頻帶判讀系統設計完成後，本論文使用此頻帶判讀系統去判讀禪定者與非禪定者的腦電波資料，經由分析判讀結果(包含：單通道頻帶統計圖與多通道腦電波頻帶時序圖)，可得到禪定者與非禪定者腦電波資料的差異性。

Abstract This thesis aims at two aspects: (1) design of a computerized interpretation system, adaptive-network-based fuzzy inference system (abbreviated as ANFIS), to classify the spectral features of EEG signals, and (2) application of the ANFIS system to the EEG signals recorded from both the experimental (meditators) and control (non-meditators) groups. The results of interpretation are demonstrated by means of running gray-level chart and statistical bar chart. Strategies of designing the ANFIS include: applying wavelet decomposition to the EEG to quantify its spectral features, partition the spectral feature space to obtain the training data set, training the ANFIS based on pre-specified criteria, and finally testing its performance. The designed ANFIS interpretation system will be used to analyze the spectral contents of the EEG in either meditation or relaxation. Perceivable differences between these two groups are observed from the results.