Deep Neural Network Architectures for Efficient Speech Emotion Recognition

View at publisher Summarise with VIRAT
Authors: Tamilarasi M, Mr. Praveen. V, Sudesh A
Cited by: 0
References: 19

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Emotions are instrumental in communication; they affect how a message is communicated and processed. This project addresses an interesting problem in communication: Speech Emotion Recognition (SER) which is the automatic identification of a person's emotional state from audio. Leveraging deep learning models for SER entails an end-to- end processing of the raw audio signal to propose predicted emotions which may include happiness, sadness, anger, fear, surprise and neutrality. The proposed model relies on using Mel-Frequency Cepstral Coefficients (MFCCs) captured from audio recordings which represent relevant features of speech signals. Two neural architectures have been proposed and evaluated; one Convolutional Neural Network (CNN) used for the local extraction of features and the second model a Bidirectional Long Short-Term Memory (BiLSTM) network to extract the temporal dependencies of speech during model training. The models are trained on labelled (emotion) dataset then evaluated their classification accuracy. The above section also discusses an additional Streamlit-based web application developed on top of the proposed system to capture userfriendliness for emotion detection. Users may either upload or record audio, and the system predicts an emotion and visualizes the probability distribution of each possible class. The proposed system indicates how deep learning-based model or models can be applied/used to either understand or interpret the affective state from audio; representing a browser application for possible and further applications in human computer interaction, call centers, mental health monitoring, and emotion-awareness virtual agents.

Fields of Study & Subjects

Emotion and Mood Recognition Speech Recognition and Synthesis Music and Audio Processing

Publication details

Journal / source
International Conference on Communication, Computing and Electronics Systems (ICCCES)
Pages
1935–1940
Publication date
21 January 2026
Language
EN