改进的蜜蜂进化型遗传算法

Improved bee evolutionary genetic algorithms

提出一种改进的蜜蜂进化型遗传算法。在该算法中,种群的最优个体作为蜂王与被选的每个个体(雄蜂)以一定概率进行交叉操作,从而增强了对种群最优个体所包含信息的开采能力;同时,为了避免过早收敛,算法在种群次优解周围进行局部搜索,引入新的随机个体,增加算法的多样性。实验结果表明,该算法能有效地提高遗传算法性能的求解精度和收敛速度。

By the use of electroencephalogram （EEG）, brain-computer interface （BCI） establishes the interfaces between human and computer. P300 is an important experiment paradigm in which the EEG signals are employed to achieve the selective text input. According to the results of detection and recognition of P300 signal in EEG, the choice of the letter given by the subject could be inferred. The dataset of P300 speller paradigm from BCI Competition III is analyzed using the algorithm designed through machine learning technique and support vector machines （SVM）. The signal channels are selected and data of fewer channels are processed. Furthermore, by adjusting the size of the dataset that participate in the training, parameter v in v-SVM obtains a larger definition interval which benefits the design of the classifier. By means of the above approaches, the classification accuracy is improved. The result on testing set revealed an accuracy of 89% for classification, which is 3% higher than the one we submitted to the competition in which linear discriminant analysis （LDA） served for the classifier.