| Summary: | 博士 === 國立成功大學 === 電機工程學系 === 103 === Meditation is used to improve psychological well-being. To enhance the efficiency of meditation practice and a meditation-induced state, it is necessary to evaluate the meditation experience using a quantitative scientific method. In this dissertation, the author reports an evaluation of meditation experience in three phases. Chapter Two compares statistical data about experienced and novice meditators. Classification of meditation experience through responses to emotional stimulation is illustrated in Chapter Three. A way to evaluate the meditation experience in real-time is described in Chapter Four.
First, statistical analysis was applied to demonstrate differences in response to emotional visual stimuli between experienced and novice meditators. The results reveal that experienced meditators showed increases in low-frequency electroencephalography (EEG) rhythms during meditation, whereas novice meditators showed increases in high-frequency EEG rhythms in response to visual stimulation. Correlational analyses show that novice meditators changed from a meditative state to a non-relaxed state when the visual stimuli were presented, whereas experienced meditators maintained the meditative state. Statistical analysis provides evidence that regular concentrative meditation can improve emotional stability, and it suggests that recording physiological responses to visual stimuli can be a good method for identifying the effects of long-term concentrative meditation.
On the basis of statistical results, artificial intelligence techniques, the support vector machine (SVM) method and the classification and regression tree (CART), were implemented to classify the three groups of meditation experiences and help validate the interaction between emotional stability and meditation experience. The results illustrate that SVM yielded a higher accuracy rate (98%) than CART (79%), and the robustness of SVM was also greater than that of CART. SVM can thus assess a meditation experience by making use of visual emotional stimulation. The results from using a data mining approach provide evidence that experienced meditators maintained calmness of mind throughout the meditation session.
To develop rapid evaluation of meditation experience, the EEG alpha responses of the participants during meditation were treated as features of the classifiers. Artificial neural networks (ANNs) and SVM were applied to evaluate the meditation experiences. Both yielded a high accuracy rate (〉 98%) in classifying meditation experiences. ANNs trained by back-propagation were stuck in local minima in only 2% of cases. The performance of SVM was highly related to feature scaling, but feature scaling had no effect on the ANN results. An extensive adjusting period and a short updated time enhanced the performance of the classifiers in evaluating the meditation experience. The artificial intelligence technologies ANN and SVM can thus be effectively used to assess the meditation experience in real time.
|
|---|
