Semantic and Affective Multimedia Retrieval using EEG-based Biological Information

在本研究中，我们利用诸如脑电图之类的生物信息来估计用户的偏好/品味，从而对图像、电影和音乐等多媒体内容进行情感信息检索。然而，当受试者看到未知内容时，很难测量脑电图。我们提出了一种基于典型相关的线性回归方法从内容特征中估计相互关联的情感特征的解决方案。作为应用，我们还提出了一种灵敏度滤波方法，用于从类似图像检索方法获得的结果中去除不愉快的图像，并证实了该方法作为情感检索系统是有效的。

In this study, we exploited methods for affective information retrieval for multimedia contents such as images, movies, and music by estimating user preferences/tastes from biological information such as EEG. It is, however, difficult to measure the EEG when a subject visualizes an unknown contents. We proposed a solution where a linear regression method based on canonical correlation is used to estimate the cross-correlation affective features from contents features. As an application, we also proposed a sensitivity filtering method for eliminating unpleasant images from the results obtained with similar image retrieval methods, and confirmed that the proposed method was efficient as an affective retrieval system.