基于中潜伏期听觉诱发电位的听觉注意特征提取与识别

Extraction and Recognition of Auditory Attention by Evoked Potentials in the Middle Latency Period

针对目前研究听觉注意的实验范式采用电极数量较多,且使用P3电位诱发时间比较长等问题,设计了一种基于听觉诱发中潜伏期反应(MLR)的实验范式。首先在注意与非注意两种状态下诱发出相应的MLR,再分别计算MLR的能量、方差、面积、AR模型系数和波形峰值作为特征值。最后,通过模式识别算法进行分类。实验结果表明:8位受试者以人工神经网络(ANN)为分类器的平均正确率可达到77.2%,本范式对于大部分受试者的效果较理想。

In view of some problems that exist in the present research on auditory attention paradigm,for example,using a large number of electrodes,longer time needs to evoke P3 potentials and so on.An experimental paradigm based on the middle latency response（MLR） is designed from auditory evoked.At first,the MLRs are respectively induced in two states of attention and non-attention; and then the energy,variance,area,AR model coefficient and waveform peak value of MLRs are respectively calculated.Finally,the features are classified by the pattern recognition algorithm.The experimental results show that the average accuracy of artificial neural network（ANN） is 77.2%from 8 subjects,which means that the paradigm can achieve satisfactory results.