CONF lathoud05b/IDIAP Multichannel Speech Enhancement in Cars: Explicit vs. Implicit Adaptation Control Lathoud, Guillaume Bourgeois, Julien Freudenberger, Jürgen EXTERNAL http://publications.idiap.ch/attachments/papers/2005/lathoud05b.pdf PUBLIC http://publications.idiap.ch/index.php/publications/showcite/lathoud-rr-04-67 Related documents Proceedings of HSCMA 2005 2005 Piscataway, NJ, USA March 2005 IDIAP-RR 04-67 Speech-based command interfaces are becoming more and more common in cars. Applications include automatic dialog systems for hands-free phone calls as well as more advanced features such as navigation systems. However, interferences, such as speech from the codriver, can significantly hamper the performance of the speech recognition component, which is crucial for those applications. This issue can be addressed with {\em adaptive} interference cancellation techniques such as the Generalized Sidelobe Canceller (GSC). In order to cancel the interference (codriver) while not cancelling the target (driver,',','), adaptation must happen only when the interference is active and dominant. This paper proposes a novel approach for pre-estimation of target and interference energies, along with its application to ``explicit'' adaptation control: a hard decision whether the filter(s) should be updated or not. It is compared with an ``implicit'' adaptation control method that does not require such pre-estimation. Experiments on real in-car recordings validate both approaches. REPORT lathoud-rr-04-67/IDIAP Sector-Based Detection for Hands-Free Speech Enhancement in Cars Lathoud, Guillaume Bourgeois, Julien Freudenberger, Jürgen EXTERNAL http://publications.idiap.ch/attachments/reports/2004/rr-04-67.pdf PUBLIC Idiap-RR-67-2004 2004 IDIAP Martigny, Switzerland Published in the EURASIP Journal on Applied Signal Processing, Special Issue on Advances in Multimicrophone Speech Processing Speech-based command interfaces are becoming more and more common in cars. Applications include automatic dialog systems for hands-free phone calls as well as more advanced features such as navigation systems. However, interferences, such as speech from the codriver, can hamper a lot the performance of the speech recognition component, which is crucial for those applications. This issue can be addressed with {\em adaptive} interference cancellation techniques such as the Generalized Sidelobe Canceller~(GSC). In order to cancel the interference (codriver) while not cancelling the target (driver,',','), adaptation must happen only when the interference is active and dominant. To that purpose, this paper proposes two efficient adaptation control methods called ``implicit'' and ``explicit''. While the ``implicit'' method is fully automatic, the ``explicit'' method relies on pre-estimation of target and interference energies. A major contribution of this paper is a direct, robust method for such pre-estimation, directly derived from sector-based detection and localization techniques. Experiments on real in-car data validate both adaptation methods, including a case with 100 km/h background road noise.