 |
This research, under funding from DARPA's ASE program, is focused on speaker authentication and related activities, noise suppression, and low-rate coding.
Speaker authentication and related activities involves a comparison of acoustic and non-acoustic measurements, glottal waveform modeling using joint frequency, improved speaker authentication, and combined speaker authentication and speech coding.
|
|
Noise suppression tasks include compensation for sensor artifacts, HMM tracking of speech and noise states, Weiner filtering, noise robust vocal tract models, and error concealment for unvoiced speech.
Low-rate coding involves coding the glottal waveform, compression based on group testing, and multi-frame interpolation of LPC (Linear Predictive Coding) models.
|
 |