单电极中潜伏期反应的听觉注意特征提取与识别  

Feature extraction and recognition of auditory attention in middle latency response from single electrode

作　　者：徐梦圆[1] 邹采荣[1,2] 梁瑞宇[1,3] 王力[2] 王青云[3] Xu Mengyuan;Zou Cairong;Liang Ruiyu;Wang Li;Wang Qingyun(School of Information Science and Engineering, Southeast University, Nanjing 210096 , China;School of Mechanica and Electric Engineering, Guangzhou University, Guangzhou 510006, China;School of Communication Engineering, Institute of Nanjing Technology, Nanjing 211167, China)

机构地区：[1]东南大学信息科学与工程学院,南京210096 [2]广州大学机械与电气工程学院,广州510006 [3]南京工程学院通信工程学院,南京211167

出　　处：《东南大学学报（自然科学版）》2017年第3期432-437,共6页Journal of Southeast University：Natural Science Edition

基　　金：国家自然科学基金资助项目(61375028;61673108);江苏省"六大人才高峰"资助项目(2016-DZXX-023);江苏省博士后科研资助计划资助项目(1601011B);江苏省"青蓝工程"资助项目;广州大学广东省灯光与声视频工程技术研究中心开放基金资助项目(KF201601;KF201602)

摘　　要：通过提取单电极中潜伏期反应(MLR)的特征差异,研究并实现了正常个体听觉注意与非注意2种状态的识别.首先,对MLR信号进行小波滤波、阈值去伪迹、相干平均等预处理;然后,分析了MLR在2种状态下的成分波差异,并将Na,Pa,Nb波的幅值与能量、面积、C0复杂度、AR模型系数等传统特征组合成为新的特征向量;最后,采用支持向量机(SVM)和人工神经网络(ANN)在传统特征向量和新特征向量下进行目标识别.8位被试的实验结果显示,在2种不同状态下,被试的Na,Pa,Nb波幅值具有显著性差异(p<0.05),而潜伏期并无差异.ANN作为分类器时,新特征向量的平均识别正确率可达85.7%.由此可见,利用单电极中潜伏期反应区分听觉注意与非注意状态是有效的.The recognition of auditory attention and non-attention states of normal individuals are studied and realized by the extraction of the differences of middle latency response(MLR)of single electrode.First,the MLR signal is preprocessed by wavelet filtering,threshold de-artifact and coherent averaging.Then,the component wave differences of these two states are analyzed.The amplitudes of the Na,Pa,Nb waves and the traditional characteristics such as the energy,the area,the CO complexity,and the coefficients of the auto regression model(AR)are combined into a new feature vector.Finally,the support vector machine(SVM)and the artificial neural network(ANN)are used to identify the target by using the traditional feature vector and the new one.The experimental results of eight subjects show that the amplitudes of the Na,Pa and Nb waves have significant differences(p<0.05)under the two different states,while no difference exhibits during the latencies.Using the new feature vector,the mean classification accuracy achieves85.7%with the ANN classifier.Therefore,it is effective to use MLR from single electrode to distinguish between the auditory attention state and the non-attention state.

关 键 词：听觉注意 中潜伏期反应 单电极 人工神经网络 

分 类 号：R318[医药卫生—生物医学工程]