K-SVM classification algorithm,combined SVM with KNN,which is adopted to identify mentality facticity,and the result is satisfactory. 15 subjects are tested respectively in mock-crime group and biographical information group. The multi-channel ERP P300 amplitude and P300 waveform area and peak to peak are used as the feature parameters and a K-SVM algorithm is applied to identify mentality facticity with the average accuracy rates are 92. 11% and 97. 37% separately. The results of both experiments show that the ac-curacy of K-SVM classification algorithm is better than pure SVM classification algorithm. Therefore,the new algorithm can provide cer-tain reference for identification of mentality facticity.%文中将支持向量机( SVM)和K近邻算法( KNN)相结合，提出一种基于K-SVM分类器的心理意识真实性识别新方法，获得了满意的结果。对15名受试者分别进行两组测试：模拟犯罪组和自传信息组。提取多通道ERP的P300幅值、波形面积和峰峰值组成特征向量，利用K-SVM算法分类，获得平均识别率分别为92.11%和97.37%，两组实验中分类精度都比单纯的SVM分类算法有明显的提高。因此文中的新算法可以为心理意识真实性检测提供一定的参考。
展开▼
