UCL logo

UCL Discovery

UCL home » Library Services » Electronic resources » UCL Discovery

Variance of spectral entropy (VSE): an SNR estimator for speech enhancement in hearing aids

Liu, F; Demosthenous, AC; Yasin, I; (2017) Variance of spectral entropy (VSE): an SNR estimator for speech enhancement in hearing aids. In: Proceedings of the 24th International Congress on Sound and Vibration. International Institute of Acoustics and Vibration (IIAV) Green open access

[img]
Preview
Text
full_paper_833_20170324132512255.pdf - ["content_typename_Accepted version" not defined]

Download (396kB) | Preview

Abstract

In everyday situations an individual can encounter a variety of acoustic environments. For an individual with a hearing aid following speech in different types of background noise can often present a challenge. For this reason, estimating the signal-to-noise ratio (SNR) is a key factor to consider in hearing-aid design. The ability to adjust a noise reduction algorithm according to the SNR could provide the flexibility required to improve speech intelligibility in varying levels of background noise. However, most of the current high-accuracy SNR estimation methods are relatively complex and may inhibit the performance of hearing aids. This study investigates the advantages of incorporating a spectral entropy method to estimate SNR for speech enhancement in hearing aids; in particular a variance of spectral entropy (VSE) measure. The VSE approach avoids some of the complex computational steps of traditional statistical-model based SNR estimation methods by only measuring the spectral entropy among frequency channels of interest within the hearing aid. For this study, the SNR was estimated using the spectral entropy method in different types of noise. The variance of the spectral entropy in a hearing-aid model with 10 peripheral frequency channels was used to measure the SNR. By measuring the variance of the spectral entropy at input SNR levels between -10 dB to 20 dB, the relationship function between the SNR and the VSE was estimated. The VSE for the speech-in-noise was measured at temporal intervals of 1.5s. The VSE method demonstrates a more reliable performance in different types of background noise, in particular for low-number of speakers babble noise when compared to the US National Institute of Standards and Technology (NIST) or Waveform Amplitude Distribution Analysis (WADA) methods. The VSE method may also reduce additional computational steps (reducing system delays) making it more appropriate for implementation in hearing aids where system delays should be minimized as much as possible.

Type: Proceedings paper
Title: Variance of spectral entropy (VSE): an SNR estimator for speech enhancement in hearing aids
Event: 24th International Congress on Sound and Vibration
Location: London
Dates: 23 July 2017 - 27 July 2017
ISBN-13: 978-1-906913-27-4
Open access status: An open access version is available from UCL Discovery
Publisher version: https://www.iiav.org/archives_icsv_last/2017_icsv2...
Language: English
Additional information: This version is the author accepted manuscript. For information on re-use, please refer to the publisher’s terms and conditions.
Keywords: Spectral entropy, signal-to-noise ratio, speech enhancement, hearing aids
UCL classification: UCL > Provost and Vice Provost Offices
UCL > Provost and Vice Provost Offices > UCL BEAMS
UCL > Provost and Vice Provost Offices > UCL BEAMS > Faculty of Engineering Science
UCL > Provost and Vice Provost Offices > UCL BEAMS > Faculty of Engineering Science > Dept of Computer Science
UCL > Provost and Vice Provost Offices > UCL BEAMS > Faculty of Engineering Science > Dept of Electronic and Electrical Eng
URI: http://discovery.ucl.ac.uk/id/eprint/1575659
Downloads since deposit
139Downloads
Download activity - last month
Download activity - last 12 months
Downloads by country - last 12 months

Archive Staff Only

View Item View Item