A BASIS METHOD FOR ROBUST ESTIMATION OF CONSTRAINED MLLR
Adaptation for ASR
Presented by: Daniel Povey, Author(s): Daniel Povey, Kaisheng Yao, Microsoft Corporation, United States
Constrained Maximum Likelihood Linear Regression (CMLLR) is a widely used speaker adaptation technique in which an affine transform of the features is estimated for each speaker. However, when the amount of speech data available is very small (e.g. a few seconds), it can be difficult to get sufficiently accurate estimates of the transform parameters. In this paper we describe a method of estimating CMLLR robustly from less data. We do this by representing the CMLLR transform matrix as a weighted sum over basis matrices, where the basis is constructed in such a way that the most important variation is concentrated in the leading coefficients. Depending on the amount of data available, we can estimate a smaller or larger number of coefficients.
Lecture Information
| Recorded: | 2011-05-24 17:55 - 18:15, Panorama |
|---|---|
| Added: | 15. 6. 2011 15:13 |
| Number of views: | 45 |
| Video resolution: | 1024x576 px, 512x288 px |
| Video length: | 0:19:14 |
| Audio track: | MP3 [6.50 MB], 0:19:14 |
Related Lectures
0:16:01
DEFINING THE CONTROLLING PARAMETER IN CONSTRAINED DISCRIMINATIVE LINEAR TRANSFORM FOR SUPERVISED SPEAKER ADAPTATION
Adaptation for ASR
Added: 15. 6. 2011 14:45
0:18:07
RAPID FEATURE SPACE MLLR SPEAKER ADAPTATION WITH BILINEAR MODELS
Adaptation for ASR
Added: 15. 6. 2011 15:27
Comments