What we do
Research in information extraction of audio and speech signals. Determining the robust representations which can succintly capture speech characteristics while being invariant to noise. Learning models which describe the multiple realizations of the data. Understanding and modeling the auditory system.
Why we do
Fundamental research in this direction to develop novel algorithms for processing audio signals. Combining signal procesing and machine learning for real world applications.
Applications
- Robust speech recognition
- Speaker Verification and Language identification
- End-to-end modeling
- EEG analysis for decoding speech perception