The automatic recognition of spontaneous emotions from speech is a challenging task. On the one hand, acoustic features need to be robust enough to capture the emotional content for various styles of speaking, and while on the other, machine learning algorithms need to be insensitive to outliers while being able to model the context. Whereas the latter has been tackled by the use of Long Short-Term Memory (LSTM) networks, the former is still under very active investigations, even though more than a decade of research has provided a large set of acoustic descriptors. In this paper, we propose a solution to the problem of 'context-aware' emotional relevant feature extraction, by combining Convolutional Neural Networks (CNNs) with LSTM networks, in order to automatically learn the best representation of the speech signal directly from the raw time representation. In this novel work on the so-called end-to-end speech emotion recognition, we show that the use of the proposed topology significantly outperforms the traditional approaches based on signal processing techniques for the prediction of spontaneous and natural emotions on the RECOLA database.

Abstract-We present in this paper a new multimodal corpus of spontaneous collaborative and affective interactions in French: RECOLA, which is being made available to the research community. Participants were recorded in dyads during a video conference while completing a task requiring collaboration. Different multimodal data, i.e., audio, video, ECG and EDA, were recorded continuously and synchronously. In total, 46 participants took part in the test, for which the first 5 minutes of interaction were kept to ease annotation. In addition to these recordings, 6 annotators measured emotion continuously on two dimensions: arousal and valence, as well as social behavior labels on five dimensions. The corpus allowed us to take self-report measures of users during task completion. Methodologies and issues related to affective corpus construction are briefly reviewed in this paper. We further detail how the corpus was constructed, i.e., participants, procedure and task, the multimodal recording setup, the annotation of data and some analysis of the quality of these annotations.

baseline systems on the three proposed tasks: state-of-mind recognition, depression assessment with AI, and cross-cultural affect sensing, respectively.

No abstract