Research project

Voice loading and speech intelligibility in different workplace environments 

Investigating the vocal habits of people who have noisy workplaces is problematic. In a quiet clinic, it is not easy to find out what the patient really does with her voice when confronted with the real noise. Making recordings on location, on the other hand, is not only cumbersome but also of limited value for detailed analysis, since the recordings will be contaminated by the offending noise itself. Voice loading studies are often done using headphones, which disrupt the auditory feedback and are difficult to calibrate for level. In this project we develop a novel method that addresses these problems.
The noisy environment is faithfully re-created or simulated in the laboratory, over high fidelity loudspeakers. The subject is asked to attempt vocal communication in the simulated noise, and her voice is recorded together with the noise. This recording can be so noisy that it is unintelligible. However, because the noise signal is now known beforehand, it is possible to use so-called adaptive filtering to cancel out the noise and thus recover the voice signal alone. The amount of noise rejection that can be achieved is proportional to the precision of the experimental setup, and to the volume of number-crunching that is subsequently invested. This signal processing is normally done off-line, but even with a setup that runs in real time, noise rejection of up to 20 dB is possible over a useful frequency range. With off-line processing, up to 40 dB noise rejection can be achieved. Currently, the limiting factor is the (unfortunate) need for the subject to remain motionless so as not to disturb the acoustic channel. The noise signal can be any appropriate sound, such as hi-fi recordings of actual workplace noise.
This method makes it easier for the clinician not only to provoke a realistic sample of voice abuse, but also to acquire a recording that is clean enough for spectral analysis, to play it back to the patient, and to explore alternative vocal strategies for workplaces where the noise cannot be abated. Pilot experiments have been completed and the method has been shown to work. Measurements of the phonatory habits of normal speakers in noise are currently under way.

Project leader: Sten Ternström, Ph.D.
Project participants: Maria Södersten, Ph.D.; Mikael Bohman,
Funded by FAS, contract #2001-0341

My homepage | Music Acoustics | Music group | Research of the group

KTH - Royal Institute of Technology
TMH - Department of Speech, Music and Hearing
Drottning Kristinas v. 31 SE-100 44 Stockholm, Sweden

phone +46 (8) 790 78 55
fax +46 (8) 790 78 54

Updated 2001.09.21 Maintained by: