fMPE: Discriminatively Trained Features for Speech Recognition
Lists
Only you see this
Abstract
MPE (minimum phone error) is a previously introduced technique for discriminative training of HMM parameters. fMPE applies the same objective function to the features, transforming the data with a kernel-like method and training millions of parameters, comparable to the size of the acoustic model. Despite the large number of parameters, fMPE is robust to over-training. The method is to train a matrix projecting from posteriors of Gaussians to a normal size feature space, and then to add the projected features to normal features such as PLP. The matrix is trained from a zero start using a linear method. Sparsity of posteriors ensures speed in both training and test time. The technique gives similar improvements to MPE (around 10% relative). MPE on top of fMPE results in error rates up to 6.5% relative better than MPE alone, or more if multiple layers of transform are trained.
BibTeX
@inproceedings{povey2005fmpe,
author = {Daniel Povey and Brian Kingsbury and Lidia Mangu and George Saon and Hagen Soltau and Geoffrey Zweig},
title = {{fMPE: Discriminatively Trained Features for Speech Recognition}},
booktitle = {IEEE International Conference on Acoustics, Speech and Signal Processing},
year = {2005},
url = {https://openreview.net/forum?id=HF8awZ0ALg5},
}Loading related content...