Multimodal Sentiment Analysis based on Video and Audio Inputs

TL;DR

This paper explores multimodal sentiment analysis using video and audio inputs, employing emotion recognition models and various decision frameworks to improve accuracy, demonstrating promising results for future research.

Contribution

It demonstrates the usability of emotion recognition models on video and audio data for sentiment analysis and introduces multiple decision frameworks to enhance accuracy.

Findings

Emotion recognition models can effectively analyze sentiment from video and audio.

Decision frameworks like Weighted Average and Dynamic Weighting improve accuracy.

Encouraging results suggest potential for further research in multimodal sentiment analysis.

Abstract

Despite the abundance of current researches working on the sentiment analysis from videos and audios, finding the best model that gives the highest accuracy rate is still considered a challenge for researchers in this field. The main objective of this paper is to prove the usability of emotion recognition models that take video and audio inputs. The datasets used to train the models are the CREMA-D dataset for audio and the RAVDESS dataset for video. The fine-tuned models that been used are: Facebook/wav2vec2-large for audio and the Google/vivit-b-16x2-kinetics400 for video. The avarage of the probabilities for each emotion generated by the two previous models is utilized in the decision making framework. After disparity in the results, if one of the models gets much higher accuracy, another test framework is created. The methods used are the Weighted Average method, the Confidence Level Threshold method, the Dynamic Weighting Based on Confidence method, and the Rule-Based Logic method. This limited approach gives encouraging results that make future research into these methods viable.