Classification of the emotional impact initiated by film sequencesAssessment of the emotional experience of a multimedia presentation consisting of video and audio.

体验质量这个概念试图评估观察者在消费多媒体内容时所体验到的主观质量。在目前的建议中，QOE应该通过一个新的中心组件来扩展，即观察者在观看视频时所经历的情绪状态。该项目的核心是利用多种机器学习方法提取视频中与情感相关的关键特征，然后用这些特征来预测观察者的经验情感。为了从技术上实现这一跨学科研究问题，我们借鉴和扩展了情感心理模型，并由工程师计算实现。有监督的机器学习方法是通过比较预测的情感体验和观察者的实际情感体验来训练的。这个学习阶段将使用众包方法，允许以相对较低的成本收集大量数据。大样本允许对结果进行广泛的概括，并允许调查观察者的个性特征如何调节视频的情绪影响。在线研究的结果将在受控实验室实验中进行交叉验证和扩展。在这项研究中，客观和间接的情绪指标被评估(面部表情的肌电图；来自观察者的视频记录的(微)情绪的编码)。这个跨学科的项目由来自心理学和计算机工程师的研究人员组成，并致力于以下研究问题：A)是否有可能确定技术特征到情感相关刺激的映射？B)心理模型是否能够根据视频的相关刺激和人的个性特征正确地预测体验到的情绪？该项目的目标一方面是实现现有的情绪心理学理论，使用大型数据集验证它们，并在这些结果的基础上进一步发展它们。另一方面，应该利用技术框架对大量涉及对观察者的情感影响的视频进行分类。此外，该项目将生成一个标准化的视频数据库，其情感影响是众所周知的。我们预计，这样的数据库将对未来人类情感领域的研究产生巨大的影响和好处。

The concept Quality of Experience (QoE) tries to assess the subjective quality which an observer experiences when consuming multimedia content. In the current proposal, QoE should be extended by a new central component, namely the emotional state an observer experiences when watching a video. The core of our project is to extract emotionally relevant key features if videos using several machine learning approaches, which in turn are used to predict the experienced emotion of the observer.In order to realize this interdisciplinary research question technically, we draw upon and extend psychological models of emotion, which are then computationally implemented by engineers. Supervised machine learning approaches are trained by comparing the predicted with the actual emotional experience of an observer. This learning stage will use a crowdsourcing approach which allows to gather a large amount of data for relatively low costs. The large sample allows a broad generalizability of the results, and allows to investigate how personality characteristics of the observers modulate the emotional impact of the videos. The results from the online study will be cross-validated and extended in a controlled laboratory experiment. In this study, objective and indirect indicators of the experienced emotions are assessed (electromyography of facial expressions; codings of (micro-)emotions from video recordings of the observers). The interdisciplinary project is composed of researchers from psychology and computer engineers and works on the following research questions:a) Is it possible to determine a mapping of technical features onto emotionally relevant stimuli?b) Are psychological models able to correctly predict the experienced emotions based on relevant stimuli of the video and personality features of a person?Goals of the project are, on the one hand, to implement existing psychological theories of emotion, validate them using large data sets, and to develop them further based on these results. On the other hand, the technical framework should be employed to classify a large set of videos concerning the emotional impact on observers. Furthermore, the project will generate a standardized data base of videos, for which the emotional impact is very well known. We expect that such a data base will have a huge impact and benefit for future studies in the area of human emotions.