Member of the Signal Processing Group at the Institute of Telecommunications, TU Darmstadt.
SVOX Deutschland GmbH Magirus-Deutz-Str. 16
Office: S3 / 06 250
work 731 152390-481
Optimized Speech Enhancement Methods for Mobile Radio Systems
In most communication applications the recording of a speech signal takes place in a noisy environment. In case of hands-free telephony in the car the local speech signal is corrupted by background noise. Background noise typically consists of engine and wind noise as well as tire hiss.
In order to improve the speech quality, noise suppression algorithms are applied. These algorithms are typically based on spectral weightening. After a short-term spectral analysis a SNR dependent weighting factor is applied to the input spectra. The enhanced speech signal results from the recombined weighted spectra. Up to now, most methods do not consider the (speech-)coding and decoding that is necessary for the transmission to the remote speaker. Therefore, the quality of a noise suppression algorithm and in particular the quality differences between specific algorithms could be affected.
The objective of this work is to investigate and to implement new noise suppression algorithms. Alongside the influence of current coding and decoding methods should be considered.