Improving speech perception in noise for people with hearing loss by using auditory-inspired machine learning. (CLaS-CCD Research Colloquium Series)
Speaker : Mr Tobias Goehring, Engineering and the Environment, University of Southampton, UK.
Date : 20th of March 2017, 11:00AM until 12:00PM
Location : E6A, Level 3, Room 357, Macquarie University.
Speech understanding in noisy environments is still one of the major challenges for users of hearing aids (HA) and cochlear implants (CI) in everyday life. We will give a quick overview of the processing paradigms and technical challenges for hearing devices and the perceptual consequences that people with hearing loss face everyday. Furthermore, we present a speech enhancement algorithm based on neural networks (NNSE) for improving speech intelligibility in noise for HA and CI users. The algorithm decomposes the noisy speech signal into time-frequency units, extracts a set of auditory-inspired features and feeds them to the neural network to produce an estimation of which frequency channels contain more perceptually important information (higher signal-to-noise ratio, SNR). The proposed algorithm was evaluated by measuring the speech-in-noise performance of 14 CI users using three types of background noise. The results indicate that the algorithm has the potential to improve the intelligibility of speech in noise for both HA and CI users while meeting the technical requirements of low computational complexity and processing delay for application in hearing devices. We aim to conclude the presentation with an open discussion on potential venues for improvement of the machine learning part of the algorithm.
Upcoming CCD Seminars
- Wednesday 29th Mar,
Professor Ocke-Schwen Bohn,
"Second language speech learning: Do cross-language phonetic ..."
- Friday 31st Mar,
Professor Paula Fikkert,
"Umlaut in the history of West Germanic with particular focus on Dutch. ..."
- Wednesday 12th Apr,
Dr Danielle Colenbrander,
"Morphological instruction for children with reading and spelling ..."
- Wednesday 19th Apr,
"Beginner guide to Magnetoencephalography (MEG)"