基于小波预处理和贝叶斯分类器的P300识别算法

P300 Detection Algorithm Based on Wavelet Preprocessing and Bayesian Classification

提出了一种高效的诱发电位P300成分识别算法用于脑计算机接口。采用小波分解与重构法去噪,根据P300特征决定小波基函数和分解层数,抽取出最明显的特征成分,结合基于证据框架的贝叶斯回归学习方法,获得对应类别概率进行分类决策。数据来源于2004 BCI Competition中的dataset P300字符拼写实验,交叉验证的结果表明,滤波方法有效,特征提取和分类算法计算复杂度低,获得了比较高的分类精度,平均精度最高为90%。

Effective algorithm for P300 detection is presented to implement brain computer interface. Firstly, wavelet decomposition and reconstruction are applied to remove noise, wavelet basis function and decomposition level are chosen by P300 character, and feature component is extracted distinctly. Secondly, it obtains the corresponding classification probability and implements classification discriminant with Bayesian regression learning algorithm on evidence framework. Data are from P300 speller paradigm in dataset Ⅱ of 2004 BCI Competition Ⅲ. Cross validation results indicate that the filtering method is effective, the computation complexity of feature extraction and classification is low, and better classification precision is ob- tained for average 90%.