This work reviews the state of the art in multimodal speech emotion recognition methodologies, focusing on audio, text and visual information. We provide a new, descriptive categorization of methods, based on the way they handle the inter-modality and intra-modality dynamics in the temporal dimension: (i) non-temporal architectures (NTA), which do not significantly model the temporal dimension in both unimodal and multimodal interaction; (ii) pseudo-temporal architectures (PTA), which also assume an oversimplification of the temporal dimension, although in one of the unimodal or multimodal interactions; and (iii) temporal architectures (TA), which try to capture both unimodal and cross-modal temporal dependencies. In addition, we review the basic feature representation methods for each modality, and we present aggregated evaluation results on the reported methodologies. Finally, we conclude this work with an in-depth analysis of the future challenges related to validation procedures, representation learning and method robustness.

The exponential growth of user-generated content has increased the need for efficient video summarization schemes. However, most approaches underestimate the power of aural features, while they are designed to work mainly on commercial/professional videos. In this work, we present an approach that uses both aural and visual features in order to create video summaries from user-generated videos. Our approach produces dynamic video summaries, that is, comprising the most “important” parts of the original video, which are arranged so as to preserve their temporal order. We use supervised knowledge from both the aforementioned modalities and train a binary classifier, which learns to recognize the important parts of videos. Moreover, we present a novel user-generated dataset which contains videos from several categories. Every 1 sec part of each video from our dataset has been annotated by more than three annotators as being important or not. We evaluate our approach using several classification strategies based on audio, video and fused features. Our experimental results illustrate the potential of our approach.

Multimodal Language Analysis is a demanding area of research, since it is associated with two requirements: combining different modalities and capturing temporal information. During the last years, several works have been proposed in the area, mostly centered around supervised learning in downstream tasks. In this paper we propose extracting unsupervised Multimodal Language representations that are universal and can be applied to different tasks. Towards this end, we map the word-level aligned multimodal sequences to 2-D matrices and then use Convolutional Autoencoders to learn embeddings by combining multiple datasets. Extensive experimentation on Sentiment Analysis (MOSEI) and Emotion Recognition (IEMOCAP) indicate that the learned representations can achieve near-state-of-the-art performance with just the use of a Logistic Regression algorithm for downstream classification. It is also shown that our method is extremely lightweight and can be easily generalized to other tasks and unseen data with small performance drop and almost the same number of parameters. The proposed multimodal representation models are open-sourced and will help grow the applicability of Multimodal Language.

No abstract

No abstract