Method and system for reducing a voice signal noise
View Patent ↗A method is provided whereby, before being subjected to a low rate voice coding, an incoming digital voice signal is chronologically segmented into blocks, the blocks are broken down respectively, in chronological order, into frequency components by a transformation in the frequency range and the frequency components are multiplied by weight factors depending on the frequency and modifiable in time, a frequency component being multiplied by the last weight factor calculated for the frequency component if the factor is less than the current weight factor.
1. A method for voice processing, comprising:
segmenting an incoming digital voice signal chronologically into blocks;
mapping the blocks in chronological order, by a transformation in a respective frequency range, onto respective frequency components;
multiplying the frequency components by chronologically modifiable frequency-dependent weighting factors derived from estimated a-priori and a-posteriori signal-to-noise ratios having a plurality of values, wherein:
a respective frequency component is multiplied by a current weighting factor if the current weighting factor is smaller than a weighting factor last calculated for the frequency component, and
the frequency component is multiplied by the weighting factor last calculated for the frequency component if the weighting factor last calculated is smaller than the current weighting factor, and
feeding the weighted frequency components back, after a back transformation in a respective time range, to a low-rate voice codec; and wherein
the a-priori signal-to-noise ratio is defined as a power density spectrum of the incoming digital voice signal and an a-priori noise estimation, and
the a-posteriori signal-to-noise ration is defined as the power density spectrum of the incoming digital voice signal and an output signal of a buffering.
2. A method for voice processing as claimed in claim 1 , wherein a respective frequency component is multiplied by the current weighting factor if the respective frequency-dependent weighting factor lies above a threshold value.
3. A method for voice processing as claimed in claim 1 , wherein a respective frequency component is multiplied by the current weighting factor if the current weighting factor lies above a threshold value, and if the weighting factor last calculated for the frequency component is smaller than the current weighting factor.
4. A system for noise suppression, comprising:
an input for digital voice signals; and
a processor unit for chronologically segmenting an incoming digital voice signal into blocks, for mapping the blocks in chronological order, by a transformation in a respective frequency range, onto respective frequency components, for multiplying the frequency components by chronologically modifiable frequency-dependent weighting factors derived from estimated a-priori and a-posteriori signal-to-noise ratios having a plurality of values, wherein
a respective frequency component is multiplied by a current weighting factor if the current weighting factor is smaller than a weighting factor last calculated for the frequency components, the frequency component is multiplied by the weighting factor last calculated for the frequency component if the weighting factor last calculated is smaller than the current weighting factor, and for feeding the weighted frequency components back, after a back transformation in a respective time range, to a low-rate voice codec,
the a-priori signal-to-noise ratio is defined as a power density spectrum of the incoming digital voice signal and an a-priori noise estimation, and
the a-posteriori signal-to-noise ration is defined as the power density spectrum of the incoming digital voice signal and an output signal of a buffering.
5. A system for noise suppression as claimed in claim 4 , wherein a respective frequency component is multiplied by the current weighting factor if the respective frequency-dependent weighting factor lies above a threshold value.
6. A system for noise suppression as claimed in claim 4 , wherein a respective frequency component is multiplied by the current weighting factor if the current weighting factor lies above a threshold value, and if the weighting factor last calculated for the frequency component is smaller than the current weighting factor.