再认记忆的事件相关电位多域特征值研究

Multi-Domain Feature of Event-Related Potential in Recognition Memory

为了更好地了解再认记忆状态下的事件相关电位(ERP)在时域、频域和空间域的性质,与传统的ERP信号特征提取方法往往都局限于时域特征不同,该研究采用非负张量分解(NTF)技术,提取再认记忆实验中与"Old"和"New"刺激相关的ERP的多域特征值。多域特征值是从多个导联ERP信号的时频转换中提取的,因此可以同时反映ERP在时域、频域和空间域上的性质。研究结果发现,多域特征值可以明显地反映出不同刺激类型下ERP信号的差异,与额区新旧效应相关的FN400在"New"刺激下的多域特征值显著大于"Old"刺激下的,与顶区新旧效应相关的P600在"Old"刺激下的多域特征值显著大于"New"刺激下的,说明多域特征值能很好地用于区分和识别再认记忆中的新旧刺激,这不仅为再认记忆的研究提供了一个新方法,而且为基于ERP的认知功能研究提供了新的思路和途径。

To explore the properties of event-related potential （ERP） related with recognition memory in the time, frequency and spatial domains, nonnegative tensor factorization （NTF） was applied to extract multi-domain feature from ERP signals related with ＂New＂ stimulus and ＂Old＂ stimulus respectively during recognition memory task, which is different from the traditional method to extract the feature of time domain, and the multi-domain feature was extracted from time-frequency transformation of multiple channel ERP signals so that it can reflect the properties of ERP in the time, frequency and spatial domains simultaneously. It is discovered that the multi- domain feature is able to discriminate ERP for different stimulus. The multi-domain feature of FN400 related with frontal old/new effect to ＂New＂ stimulus gets greater than the feature to ＂Old＂ stimulus. The multi-domain feature of P600 associated with parietal old/new effect to ＂Old＂ stimulus gets larger than the feature to ＂New＂ stimulus. Therefore, the multi-domain feature extracted by NTF reveals properties of ERP in the time, frequency and spatial domains, and provides a novel method to recognition memory research and cognitive function research based on ERP signal.