Spectral subtraction is a method for restoration of the spectrum magnitude for signals observed in additive noise, through subtraction of an estimate of the average noise spectrum from the noisy signal spectrum. In this paper we show that, starting from the known minimum mean-square error (MMSE) suppression rules of Ephraim and Malah and under the same modeling assumptions, a simpler suppression filtering rule can be found. Moreover, we demonstrate its performances and compare its computational costs with respect to the reference rule of Ephraim and Malah. This result permits a real time implementation of the exposed theory with an efficient algorithm on the DSP TMS320 C6713B.
A spectral subtraction rule for real-time dsp implementation of noise reduction in speech signals
ROMANIN, MATTEO;MARCHETTO, ENRICO;AVANZINI, FEDERICO
2009
Abstract
Spectral subtraction is a method for restoration of the spectrum magnitude for signals observed in additive noise, through subtraction of an estimate of the average noise spectrum from the noisy signal spectrum. In this paper we show that, starting from the known minimum mean-square error (MMSE) suppression rules of Ephraim and Malah and under the same modeling assumptions, a simpler suppression filtering rule can be found. Moreover, we demonstrate its performances and compare its computational costs with respect to the reference rule of Ephraim and Malah. This result permits a real time implementation of the exposed theory with an efficient algorithm on the DSP TMS320 C6713B.Pubblicazioni consigliate
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.