Project/Intern positions immediately available in Automatic Speech Recognition !

Latest News

CCC wav2vec2.0 is 5th on global SuPERB leaderboard

Vasista's work on developing universal speech recognition is now 5th on Global leaderboard of SuPERB benchmark. (5th in terms of Rank and 2nd in terms of overall score). The top 10 representations in Leaderboard (apart from us) are all from Microsoft or Meta-AI. We were able to achieve this by training on just 960 hours of data.

Recent Publications

Domain Adaptation of low-resource Target-Domain models using well-trained ASR Conformer Models

To Appear at IEEE SLT 2022

Sukhadia, Vrunda

PADA: Pruning Assisted Domain Adaptation For Self-Supervised Speech Representations

To Appear at IEEE SLT 2022

Lodagala, Vasista Sai;

Ghosh, Sreyan

CCC-WAV2VEC 2.0: Clustering Aided Cross Contrstive Self-Supervised Learning of Speech Representations

To Appear at IEEE SLT 2022

Lodagala, Vasista Sai;

Ghosh, Sreyan

Joint Encoder-Decoder Self-Supervised Pre-training for ASR.

Interspeech.2022-11338

Arunkumar, A

Check out our Demos

Speech to Speech

ASR

TTS

Video to Video

About Speech Lab IITM

"Automatic Speech Recognition (ASR) is the process of deriving the transcription (word sequence) of an utterance, given the speech waveform." Speech lab IIT Madras is headed by Prof. S. Umesh and is part of the Dept. of Electrical Engg. Our focus is on building state of the art speech recognition systems, especially in Indian languages. Our research interests are in low-resource modelling, multilingual speech recognition and speaker normalisation.

People at Speech Lab, IITM