Research Assistance in Bangla Speech Emotion Recognition using Emoformer

M.Sc Thesis (Research Assistance)

This work represents my contribution as a research assistant in an M.Sc thesis conducted at Pabna University of Science and Technology, under the Department of Information and Communication Engineering.

Supervisor: Prof. Dr. Md. Sarwar Hosain
Co-Supervisor: Prof. Dr. Md. Omar Faruk

The research proposes an attention-driven Emoformer architecture, integrating Convolutional Neural Networks (CNN) and Transformer models to effectively capture both local acoustic patterns and long-range temporal dependencies in speech signals.

My Contributions:

• Data preprocessing and feature extraction (MFCC, X-vectors)
• Model implementation and training
• Experimental evaluation and performance analysis

The system demonstrated strong performance across five emotional states (angry, happy, sad, surprise, neutral), highlighting its effectiveness in low-resource Bangla speech environments.

Thesis Details:
Course Code: ICE-6000
Degree: M.Sc. (Engineering) in Information and Communication Engineering
Session: 2020–2021

Prepared by:
Mst. Asma Khatun
Roll No: 21040624
Registration No: 1065220