与事件相关电位（由短时且重复刺激诱发）和自发脑电图（EEG）不同，自然刺激（比如音乐和电影等）下的EEG成像是研究人在真实经验中大脑加工的重要手段。如何处理和分析这种EEG是该研究的难点。本课题将采集健康受试者和患者在听音乐时的EEG数据，并提取音乐的长时和短时特征；然后分别在秒与分钟的时间尺度下，基于EEG的时频表述数据构建高阶张量（亦即多维数组，包含时间、频率、空间和被试维度）；进而用非负张量分解算法对高阶张量进行分解；最后选择与音乐特征显著相关的时域分量，分析该分量所对应的 EEG多域特征，从而得出结论。特别地，发展张量空间的聚类方法来判定张量分解结果的稳定性；发展基于双模态数据（音乐和EEG）潜变量间相关性最大化的联合张量分解算法，从而更有效地提取出音乐所诱发的电位；为探讨音乐治疗发挥治疗作用的脑电生理机制提供技术支持。这些工作对通过EEG进行自然刺激下的脑科学研究有学术意义。

In order to decode human brain activity during real-world experiences, ongoing EEG can be measured as one of the important approaches, instead of spontaneous EEG and event-related potentials (ERPs). Nevertheless, how to process and analyze the ongoing EEG is still an open question. In this project, when healthy adults and patients listen to pieces of naturalistic music, ongoing EEG will be collected, and the long-term and short-term musical features of the naturalistic music stimuli will be extracted. Subsequently, the time-frequency representation of ongoing EEG in different time scales will be used to construct the high-order tensor (multi-way array) including the modes of time, frequency, space and subject. Then, nonnegative tensor factorization will be applied to decompose the high-order tensor. Finally, the extracted temporal components (which are significantly correlated with the musical features) will be selected, and the corresponding multi-domain features of ongoing EEG will be analyzed for drawing conclusions. Particularly, the new tensor clustering algorithm in the tensor space will be developed to evaluate the stability of tensor decomposition results; the new coupled tensor decomposition algorithm (based on maximizing correlation of latent variables between two modalities) will be developed for more effectively extracting the evoked potentials (elicited by the naturalistic music stimuli); and the developed ongoing EEG signal processing method will assist to exploit the underlying mechanism why the musical therapy can function. The proposed research in the project is significant for the brain science research during real-world experiences.