Digital Signal Processing Reference
In-Depth Information
Figure 19-5.
Recognition performance of MRLS with optimized weights for a speaker
(SPKER), with optimized weights for a driving condition (DR), proposed weight adap-
tive method (ADAPT), with optimized weights for all training data (ALL), from left
to right.
6. SUMMARY
In this paper, we described a multichannel method of noisy speech
recognition that can adapt to various in-car noise conditions during driv-
ing. The method allows us to estimate the log spectrum of speech at
a close-talking microphone based on the multiple regression of the log
spectra (MRLS) of noisy signals captured by multiple distributed micro-
phones. Through clustering of the spatial noise distributions under var-
ious driving conditions, the regression weights for MRLS are effectively
adapted to the driving conditions. The experimental evaluation shows an
