This user has not added any information to their profile yet.
I received my B.Tech. degree in electronics and communication engineering from St. Thomas’ College of Engineering and Technology under West Bengal University of Technology, Kolkata, India. I recently submitted my M.S. thesis in the Department of Electronics and Electrical Communication Engineering, Indian Institute of Technology Kharagpur, Kharagpur, India.
My master's thesis project focused on the analysis of natural and various spoofed speech signals (voice converted, speech synthesized and playback). I thoroughly understood the current state-of-the-art anti-spoofing systems, identified the limitations, and subsequently proposed new algorithms to improve the performance of speaker verification systems under spoofing attacks.
My current research goal is to find artifact-free modification algorithms to convert from conversational to clear speech. This project will investigate various signal processing and machine learning algorithms to obtain modifications that will transform conversational speech to clear speech style effectively. The project has two aspects, the conversion of speaker identity and modification of speaking style.
Till now, the task of converting speaker identity is almost completed. I have an working voice conversion module that modifies the para/non-linguistic information contained in the speech uttered by a speaker, while keeping the linguistic contents unchanged. A conference proceeding on voice conversion has been accepted in Interspeech 2019. The proposed approach introduces an effective weight factor for each sample in the generative adversarial algorithm. Experimental results based on subjective performance evaluation confirms that our proposed method achieves better speaker similarity and perceptual speech quality than baseline systems.
Furthermore, I have also made some progress on the modification of the speaking style using Speech Recognition and Speech Synthesis framework. The idea is to implement speech recognizer to recognize the speech and later, use Speech synthesis algorithms to produce clear and well articulated speech.