Temporal-Acoustic Emotion Estimation via Deep Neural Networks

Emotion recognition from verbal communication is a critical component of affective computing, enabling more natural human-computer interactions .This paper proposes a method for temporal-acoustic emotion estimation using deep neural networks (DNNs), with a particular focus on capturing both the acoustic and timevarying features of verbal communication .By employing a Long Short-Term Memory (LSTM) network architecture, we demonstrate how incorporating temporal dynamics improves emotion recognition.Experimental results on the IEMOCAP dataset show that our approach outperforms traditional methods in emotion detection accuracy. Problem StatementExisting approaches for verbal communication-based emotion recognition largely focus on static acoustic features and often ignore the temporal variations in verbal communication.However, human emotions fluctuate over time, making it important to capture temporal dynamics in emotion estimation.This paper proposes a temporal-acoustic emotion recognition model using deep neural networks to address this challenge. ObjectiveWe aim to develop a model that effectively captures both acoustic features and their temporal evolution to improve emotion recognition performance.Specifically, we propose the use of Long Short-Term Memory (LSTM) networks, which are well-suited for modeling sequential data and capturing long-range dependencies in temporal patterns.