A Model Free Bayesian Classifier And Its Applications In Identification Of Brain Wave Signal

With the development of science and technology, for those who brain is healthy but limb can’t act, such as: amyotrophic lateral sclerosis (ALS)patients, more and more desire to communicate directly with external devices through their own brain, Brain-Computer Interface (BCI) technology appears and rapid development in this context. The classification algorithm of brainwave signal is an important part of BCI technology, and its research will have great practical significance and practical value.The main contents of this paper:1. The shortcomings of Naive Bayesian Classifier (NBC) and Bayesian Network (BN) are improved. Based on the Nearest Neighbor (NN) algorithm,a probability estimator of calculating the joint probability distribution is established, and proposed a model-free Bayesian classifier (MFBC) algorithm to achieve the classification of data.2. A new self-organizing feature extraction algorithm is proposed, and the MFBC algorithm is extended to deal with the regression problem. By comparing with the Extreme Learning Machine (ELM), the extended MFBC processing the regression problem is verified.3. The artificial brain wave signal data and the real brain wave signal data are distinguished by fuzzy C-Means (FCM) algorithm.4. The design of brainwave signal recognition system, including the import of UCI (University of California Irvine) data and brainwave data,MFBC algorithm, experimental results analysis and other functions.Finally, the sensitivity analysis of MFBC algorithm is proved by UCI data, and the convergence of MFBC algorithm is proved. The validity of the MFBC algorithm is verified by comparison with other classical and integrated classifiers. Meanwhile, the MFBC algorithm is combined with the Common Spatial Pattern (CSP) algorithm and applied in the BCI system in the signal classification procedure in order to prove its practicality.