In this paper, we examine a coding scheme for quantising feature vectors in a distributed speech recognition environment that is more robust to noise. It consists of a vector quantiser that operates on the logarithmic filterbank energies (LFBEs). Through the use of a perceptually-weighted Euclidean distance measure, which emphasises the LFBEs that represent the spectral peaks, the vector quantiser codebook provides a priori knowledge of the spectral characteristics of clean speech and is used to quantise features from noise-corrupted speech. Our comparative results from the ETSI Aurora-2 recognition task show that the perceptually-weighted vector quantisation of LFBEs achieves higher recognition accuracies for noisy speech than the unweighted vector quantisation, memoryless and multi-frame GMM-based block quantisation and scalar quantisation of Mel frequency-warped cepstral coefficients.
Cite as: So, S., Paliwal, K.K. (2005) Improved noise-robustness in distributed speech recognition via perceptually-weighted vector quantisation of filterbank energies. Proc. Interspeech 2005, 941-944, doi: 10.21437/Interspeech.2005-224
@inproceedings{so05_interspeech,
author={Stephen So and Kuldip K. Paliwal},
title={{Improved noise-robustness in distributed speech recognition via perceptually-weighted vector quantisation of filterbank energies}},
year=2005,
booktitle={Proc. Interspeech 2005},
pages={941--944},
doi={10.21437/Interspeech.2005-224}
}