News and Events

Read all the latest news within the Division of Psychology and Language Sciences.


PALS_Silver_SWAN_Small

Read more about the PALS Swan Silver Award


Divisional Subject Pool

Access and sign up for the Divisional Subject Pool.


Green Issues

The Division has won the Silver Award for the UCL Green Impact scheme


CPD@PaLS Courses

The Division runs Continuing Professional Development courses.


Perceiving speech in single and multi-talker babble in normal and impaired hearing


A research project funded by the Medical Research Council

Administrative Details
 
Grant Period
Apr 2011 - Mar 2014
Grant Award
£409,000
Grant Reference Number
G1001255
Chief Investigator
Prof Stuart Rosen
Co-Investigator and Researcher
Dr Tim Green
Researcher Dr Marine Ardoint
Collaborator Dr Pam Souza, Northwestern University


Speech is often heard in the background of other sounds, including other people talking. Listeners with normal hearing are remarkably good at filtering out extraneous sounds and listening only to the desired talker. For people with hearing impairment, however, this situation is very challenging, even though they might function very well with hearing aids in quiet. The aim of this research project is to gain a better understanding of the complex auditory processes that allow normal hearing listeners to cope so well with interfering sounds, particularly other voices, and to pinpoint how these processes are affected by hearing impairment. The information gained is expected to help guide the development of enhanced technology for hearing aids and cochlear implants.

Background

In considering how the presence of a background sound interferes with, or "masks", a target signal a distinction can be made between "energetic" and "informational" masking. Energetic masking occurs when the interfering sound overlaps in frequency and time with the signal, so that the signal is obscured at the earliest stages of auditory processing. In informational masking the signal is adequately represented at the auditory periphery, but higher levels of auditory processing are unable to segregate the signal and masker so that the signal cannot be pieced together into a coherent, intelligible whole.

One key finding with respect to energetic masking is that, for normal hearing listeners, more speech can be understood in the presence of a modulated masker (one that fluctuates in level) than an unmodulated masker of the same average intensity. This "masking release" has typically been attributed to brief periods of low energy in fluctuating maskers allowing "glimpses" of the target signal, a process also known as "listening in the dips". Hearing impaired listeners show substantially reduced masking release. Since interfering sounds in the real world, such as other voices, are typically modulated rather than stationary this lack of masking release plays an important role in the problems faced by those with impaired auditory functioning. At present, however, the reasons for this deficit are not well understood. Factors that might be important include the representation of periodicity (a regularly repeating pattern in a sound waveform, that in speech corresponds to the vibration of the vocal chords and is related to voice pitch); the representation of fine temporal details of the sound waveform; and the possibility that masking release may not occur at high signal-to-noise ratios (SNRs), which are typically required by hearing impaired listeners.

An interesting approach for investigating these issues is to measure speech recognition in the presence of varying numbers of competing voices. At one extreme, a masker consisting of a very large number of voices effectively equates to steady noise, producing exclusively energetic masking and eliminating any possibility for glimpsing. At the other extreme, with a single interfering talker, opportunities for glimpsing are maximised, but there is also likely to be substantial informational masking arising from difficulties in selecting the target rather than the interfering speech. As increasing numbers of voices are added a number of different processes will affect the masking properties of the mixture: the extent to which the background is intelligible will change, affecting the selection aspect of informational masking; the spectral and temporal dips in the masker will increasingly be filled in, reducing the opportunities for glimpsing; and the presence of additional voice pitch contours is likely to complicate the process of segregating different talkers.

Research Plans

Our main aim is to fully understand the processes underlying the performance of normal and hearing-impaired listeners (including users of cochlear implants) in understanding speech in the presence of various numbers of competing voices. We aim to elucidate factors responsible for changes in performance as the number of talkers in the background change, and what factors are responsible for differences among normal and impaired listeners. A particular focus will be on explaining the relative inability of hearing impaired listeners and cochlear implant users to effectively ‘glimpse’ speech information in fluctuating maskers, with the following possible factors being investigated: 1) the representation of periodicity in speech; 2) the role of temporal fine structure and envelope information; 3) limitations in the ability to fuse temporally-separated segments of speech across brief temporal gaps; and, 4) the effect of overall SNR on glimpsing effectiveness. Other experiments will specifically address the effects of multiple independent fundamental frequency contours in the masker and of masker intelligibility on masking effectiveness.