Improvement of Broadcast Speech Intelligibility Based on Instantaneous Listening Effort Estimate
* Presenting author
Abstract:
Depending on the hearing abilities, environment and/or age of the TV viewers, the speech intelligibility of dialogs in movies can differ markedly for individual recipients. A possibility to change the intelligibility at home on the TV set to an individual optimal level would be a desirable option for many recipients. An adaptive audio modification would be preferred because a static modification (e.g., a fixed SNR increase) may work in some scenes but be too little or too much improvement in other scenes. This contribution therefore presents a way to provide a second, automatically realized version of a speech/background mix for people with increased demands on intelligibility. The degree of intervention depends on the instantaneously estimated listening effort in relation to the target or individually desired listening effort. To this end, a single-ended (i.e. reference-free) method is used to evaluate the instantaneous listening effort and to inform a subsequent stage of blind source separation and dialog enhancement. The latter can be configured to a target listening effort, which is not to be exceeded, and can hence be adjusted to improve the speech intelligibility of broadcast audio for individual needs or different target groups.