9th Annual Conference of the International Speech Communication Association

Brisbane, Australia
September 22-26, 2008

Time and Frequency Dependent Amplification for Speech Intelligibility Enhancement in Noisy Environments

Henk Brouckxon (1), Werner Verhelst (1), Bart De Schuymer (2)

(1) Vrije Universiteit Brussel, Belgium; (2) Televic nv, Belgium

When speech is presented through loudspeakers in a noisy environment, the background noise can significantly decrease speech intelligibility. Because the amplitude and spectrum of the background noise can vary over time (and because high loudness levels are to be avoided for listener comfort), choosing proper speech equalization and master gain settings for a public address system can be a difficult task. In this paper, we propose an adaptive digital signal processing algorithm that applies a frequency and time dependent gain strategy to the speech signal in order to enhance its intelligibility in noise with a minimal increase of the overall sound energy level. An alternative version of the system can also be used to maximise speech intelligibility without increasing the overall energy level of the signal. The proposed algorithm makes use of the psycho-acoustic masking properties of the human hearing system and relies on the importance of the formant information for speech intelligibility.

Full Paper

Bibliographic reference.  Brouckxon, Henk / Verhelst, Werner / Schuymer, Bart De (2008): "Time and frequency dependent amplification for speech intelligibility enhancement in noisy environments", In INTERSPEECH-2008, 557-560.