Education
University of Southern Queensland, Australia
October 2017 - April 2021Doctor of Philosophy in Information Technology
Thesis: Learning Unsupervised Disentangled Representation From Audio for Transfer Learning
Jahangirnagar University, Bangladesh
May 2015 – September 2016Master in Information Technology
Khulna University of Engineering & Technology, Bangladesh
February 2010 – February 2014Bachelor of Science in Urban and Regional Planning
Projects
Speech Emotion Recognition
I received a highly competitive International Scholarship to pursue my PhD, which is funded through the Advance Queensland (AQ) Fellowship Project of Dr Rana. The project aims to develop an automated Mood Inference Tool to detect mood changes from phone calls. The core building block for Mood Inference Tool is emotion detection from spontaneous speech, where the accuracy in the literature for emotion recognition is very low. Under Dr Rana’s instructions, I have developed a novel “Deep Neural Network” based system, which can detect emotions with 98% accuracy, where the current state of the art accuracy is 72%. This is a breakthrough. I have implemented the mood inference system in a complex cloud environment, which was used for trials with mental health patients at the Royal Brisbane Mental Health Services. I have also assisted the Industry partner in integrating the mood inference system in their web-based platform Nexusonline.
View Project
Voice Tone Prediction
I have a built a voice tone analysis demo to analyse the voice tone of any human from audio data. The model is built mixing bleeding-edge Deep Learning models.
View ProjectResearch
Guided Generative Adversarial Neural Network for Representation Learning and Audio Generationusing Fewer Labelled Audio Data
View Details
High-Fidelity Audio Generation and Representation Learning With Guided Adversarial Autoencoder
View Details
Image denoising and restoration with CNN-LSTM Encoder Decoder with Direct Attention
View Details