基于AAR模型的听觉诱发中潜伏期反应特征提取  

Feature extraction of the middle latency response in auditory based on AAR model

作　　者：董倩妍 王力[1] 蒋本聪 胡晓[1] Dong Qianyan Wang Li Jiang Bencong Hu Xiao(School of Mechanical and Electrical Engineering,Guangzhou University,Guangzhou 510006, China)

机构地区：[1]广州大学机械与电气工程学院,广东广州510006

出　　处：《电子技术应用》2017年第11期78-81,共4页Application of Electronic Technique

基　　金：广州市科技计划项目(201605030014);广州市市属高校科研项目(1201630210);广州大学科技创新培育基金

摘　　要：针对听觉刺激诱发的脑干中潜伏期反应(MLR)信号的非平稳特性,采取计算其自适应自回归(AAR)模型系数的方法进行特征提取,实现不同注意状态的分类。首先对采集的MLR数据进行去噪预处理,然后结合相对误差方差(REV)准则分别进行卡尔曼滤波和最小均方误差自适应算法估计其AAR模型参数。利用支持向量机对两种估计方法的特征参数分别进行分类。最后根据最大互信息和分类结果进行比较,最小均方误差自适应算法估计AAR模型系数的分类正确率达到77.45%,最大互信息值为0.3011,其效果优于卡尔曼滤波算法。Aiming at the non-stationarity of the middle latency response（MLR） brain stem signals, the adaptive autoregressive model （AAR） coefficients is used to extract the features, which is to achieve the classification of different attention states. Firstly, the ex- tracted MLR data is pretreated to remove noises. Then, the Kalman filter and the least mean square error adaptive estimation method are used to extract the AAR model parameters respectively according to the REV criterion. The support vector machine is used to classify the characteristic parameters of the two estimation methods. Finally, based on the comparison of maximum mutual information and classification results, the minimum mean square error adaptive algorithm estimates that the classification accuracy of AAR model coefficients is 77.45% and the maximum mutual information value is 0.3011, which is better than Kalman filter algorithm.

关 键 词：卡尔曼滤波 最小均方误差自适应 相对误差方差准则 支持向量机 互信息 

分 类 号：TN911.7[电子电信—通信与信息系统]