New research to enhance speech recognition technology

New research is hoping to understand how the human brain hears sound to help develop improved hearing aids and automatic speech recognition systems.

Led by the Universities of Southampton and Cambridge, the research aims to develop physiologically-inspired algorithms, which mimic how our brain hears sound to improve on traditional algorithms.

The novelty of the research is that instead of looking at signal energy – as today's artificial devices do – the researchers are concentrating on how the brain processes sound information instead.

Dr Stefan Bleeck, from the Institute of Sound and Vibration Research at the University of Southampton, is looking to create algorithms based on neuronal responses to give insights into how sound is coded within the brain. Once researchers know better how sound is coded, they will be able to select the parts that code speech and the ones that code unwanted noise. They will then be able to resynthesize sound in aids with reduced noise, but with quality intact, to enhance speech intelligibility.

Today's speech enhancement systems can reduce noise and increase speech quality, but they are not good at improving speech intelligibility, especially in noisy situations where users have to concentrate to pick out single speakers. With about 10 per cent of the UK population hearing impaired, current signal processing technology hasn't come up with a suitable system to enhance speech intelligibility.

Dr Bleeck says: "Today, it is still the ultimate goal for the speech signal processing community to develop speech enhancement systems that perform as well as humans in noisy situations. Normal hearing humans still easily outperform any technical system – sound processing in the brain is more successful than signal processing in silicone. A system that works as well as a human would lead to the next revolution in human communication and would greatly benefit hearing impaired people.

"My vision is to build a brain-inspired speech enhancer in the next five years, which will be able to identify sources and to enhance intelligibility. This should be useful in everyday situations, for hearing impaired as well as normal hearing people, so that it ultimately reduces the stigma that have today. Using this device in the future to hear better should be as normal as wearing glasses today to see better."

Dr Bleeck has received funding from a Google Research Award to undertake this research with ISVR colleague Dr Matthew Wright, and Dr Ian Winter of the University of Cambridge's Department of Physiology.

Related Stories

Bionic speech recognition

date Sep 09, 2010

As speech recognition systems become more commonplace - on the computer desktop top, at the call centre and even in the car - it is increasingly important to ensure that the voice signal is as clear as possible before it ...

Hearing the words beneath the noise

date Aug 05, 2009

Hearing aids and cochlear implants act as tiny amplifiers so the deaf and hard-of-hearing can make sense of voices and music. Unfortunately, these devices also amplify background sound, so they're less effective in a noisy ...

Recommended for you

Making waves with groundbreaking brain research

date Jul 03, 2015

New research by Jason Gallivan and Randy Flanagan suggests that when deciding which of several possible actions to perform, the human brain plans multiple actions simultaneously prior to selecting one of ...

Long-term memories are maintained by prion-like proteins

date Jul 02, 2015

Research from Eric Kandel's lab at Columbia University Medical Center (CUMC) has uncovered further evidence of a system in the brain that persistently maintains memories for long periods of time. And paradoxically, ...

Water to understand the brain

date Jul 02, 2015

To observe the brain in action, scientists and physicians use imaging techniques, among which functional magnetic resonance imaging (fMRI) is the best known. These techniques are not based on direct observations ...

User comments

Please sign in to add a comment. Registration is free, and takes less than a minute. Read more

Click here to reset your password.
Sign in to get notified via email when new comments are made.