Edinburgh Research Explorer

Intelligibility enhancement of HMM-generated speech in additive noise by modifying Mel cepstral coefficients to increase the glimpse proportion

Research output: Contribution to journalArticle

Related Edinburgh Organisations

Documents

  • Download as Adobe PDF

    Rights statement: © Valentini-Botinhao, C., Yamagishi, J., King, S., & Maia, R. (2014). Intelligibility enhancement of HMM-generated speech in additive noise by modifying Mel cepstral coefficients to increase the glimpse proportion. Computer Speech & Language, 28(2), 665-686. 10.1016/j.csl.2013.06.001

    Submitted manuscript, 2 MB, PDF document

Original languageEnglish
Pages (from-to)665-686
Number of pages22
JournalComputer Speech and Language
Volume28
Issue number2
DOIs
Publication statusPublished - Mar 2014

Abstract

This paper describes speech intelligibility enhancement for Hidden Markov Model (HMM) generated synthetic speech in noise. We present a method for modifying the Mel cepstral coefficients generated by statistical parametric models that have been trained on plain speech. We update these coefficients such that the glimpse proportion - an objective measure of the intelligibility of speech in noise - increases, while keeping the speech energy fixed. An acoustic analysis reveals that the modified speech is boosted in the region 1-4 kHz, particularly for vowels, nasals and approximants. Results from listening tests employing speech-shaped noise show that the modified speech is as intelligible as a synthetic voice trained on plain speech whose duration, Mel cepstral coefficients and excitation signal parameters have been adapted to Lombard speech from the same speaker. Our proposed method does not require these additional recordings of Lombard speech. In the presence of a competing talker, both modification and adaptation of spectral coefficients give more modest gains. (C) 2013 Elsevier Ltd. All rights reserved.

    Research areas

  • Intelligibility of speech in noise, HMM-based speech synthesis, Mel cepstral coefficients, Glimpse proportion measure, ALGORITHMS, HEARING, MODEL

Download statistics

No data available

ID: 14407515