DAGA 2010 - Deutsche Gesellschaft für Akustik eV

94 DAGA 2010 Programm
Di. 15:45 Bauwesen H1 Sprache im Kraftfahrzeug
Head-Unit Integrated Microphone Array for Handsfree Wideband
Telephony
H. Yu und T. Fingscheidt
TU Braunschweig, Institut für Nachrichtentechnik
Nowadays handsfree equipment for mobile telephony is mandatory in
cars. With upcoming wideband telephony appropriate solutions operating
at a sampling frequency of 16 kHz are required. Microphone array
based beamforming for noise reduction is particularly interesting in car
applications, however, it is challenging. In this paper we propose a headunit
integrated microphone array with low-cost microphones. Unlike many
other approaches assuming quite idealistic conditions, our proposal
is optimized and tested with real multi-channel signals acquired with these
integrated microphones. Although the position of the head-unit is not
optimal from an acoustical perspective, it is yet very attractive, since no
further wiring is necessary and radio/navigation systems manufacturers
can offer very compact and optimized solutions including the microphones.
In order to achieve good noise reduction performance in car noise,
we complement the beamformer with a post-filter, which is estimated
with the knowledge of the car specific noise coherence of diffuse noise
and an adaptive smoothing approach. By using an intrusive instrumental
evaluation methodology, we show that compared to the state of the art
a significant level of noise attenuation along with a high preservation of
the quality of the speech component can be achieved.
Di. 16:35 Bauwesen H1 Sprache im Kraftfahrzeug
Robust adaptive cancellation of interfering speakers for distributed
microphone systems in cars
T. Matheja, M. Buck und T. Wolff
Nuance Communications
For hands-free speaking systems in car environments often multichannel
signal processing is applied to include the speech of the wanted
speaker at the best. Beside beamforming approaches distributed microphone
systems can be installed where the distances between the microphones
might be larger in order to align to several potential speakers.
In this presentation a system is proposed that applies adaptive filters
to cancel cross-talk between the microphone signals. In contrast to the
well-known echo cancellation problem a noise component occurs on the
reference signal which complicates the control of the adaptive filter. A
robust adaptation rule based on the NLMS algorithm is derived and discussed.