Enhanced de-esser for in-car communication systems
Methods and systems for deessing of speech signals are described. A deesser of a speech processing system includes an analyzer configured to receive a full spectral envelope for each time frame of a speech signal presented to the speech processing system, and to analyze the full spectral envelope to identify frequency content for deessing. The deesser also includes a compressor configured to receive results from the analyzer and to spectrally weight the speech signal as a function of results of the analyzer. The analyzer can be configured to calculate a psychoacoustic measure from the full spectral envelope, and may be further configured to detect sibilant sounds of the speech signal using the psychoacoustic measure. The psychoacoustic measure can include, for example, a measure of sharpness, and the analyzer may be further configured to calculate deesser weights based on the measure of sharpness. An example application includes in-car communications.
1. A method of deessing a speech signal, the method comprising
for each time frame of a speech signal presented to a speech processing system, analyzing a full spectral envelope to identify frequency content for deessing and
spectrally weighting the speech signal as a function of results of the analyzing
wherein spectrally weighting the speech signal includes applying deesser weights to sibilant sounds of the speech signal,
wherein the method further comprises applying the deesser weights to control attack and release of a compressor, and
wherein the compressor includes a soft threshold and a hard threshold, the soft threshold causing the compressor to moderate the further increase in a measure of sharpness for a given ratio R, the hard threshold being a not-to-exceed threshold of the measure of sharpness.
2. The method of claim 1 , wherein spectrally weighting the speech signals occurs in the frequency domain and at a frequency resolution matching that of the full spectral envelope.
3. The method of claim 1 , wherein analyzing the full spectral envelope includes calculating a psychoacoustic measure from the full spectral envelope.
4. The method of claim 3 , wherein analyzing the full spectral envelope further includes detecting sibilant sounds of the speech signal using the psychoacoustic measure.
5. The method of claim 3 , wherein the psychoacoustic measure includes at least one of a measure of sharpness and a measure of roughness.
6. The method of claim 3 , wherein the psychoacoustic measure includes a measure of sharpness, and wherein analyzing the full spectral envelope further includes calculating deesser weights based on the measure of sharpness.
7. The method of claim 1 , further comprising calculating the measure of sharpness including calculating a measure of sharpness without application of the deesser weights and calculating another measure of sharpness with application of the deesser weights.
8. The method of claim 7 , wherein controlling the attack and release of the compressor includes, (i) if the measure of sharpness calculated with application of the deesser weights exceeds one of the thresholds of the compressor, adapting the deesser weights according to a gradient-descent method to attack those parts of the spectral envelope that dominate the measure of sharpness, otherwise, (ii) releasing the deesser weights.
9. A deesser of a speech processing system, the deesser comprising an analyzer and a compressor,
wherein the analyzer is configured to receive a full spectral envelope for each time frame of a speech signal presented to the speech processing system and to analyze the full spectral envelope to identify frequency content for deessing
wherein the compressor is configured to receive results from the analyzer and to spectrally weight the speech signal as a function of results of the analyzer,
wherein the compressor is configured to spectrally weight the speech signal by applying deesser weights to sibilant sounds of the speech signal, and
wherein the compressor includes a soft threshold and a hard threshold, the soft threshold causing the compressor to moderate the further increase in a measure of sharpness for a given ratio R, the hard threshold being a not-to-exceed threshold of the measure of sharpness.
10. The deesser of claim 9 , wherein the analyzer is configured to calculate a psychoacoustic measure from the full spectral envelope.
11. The deesser of claim 10 , wherein the analyzer is further configured to detect sibilant sounds of the speech signal using the psychoacoustic measure.
12. The deesser of claim 10 , wherein the psychoacoustic measure includes a measure of sharpness, and wherein the analyzer is further configured to calculate deesser weights based on the measure of sharpness.
13. The deesser of claim 9 , wherein the analyzer is configured to calculate at least two measures of sharpness, a measure of sharpness without application of the deesser weights and another measure of sharpness with application of the deesser weights.
14. The deesser of claim 13 , wherein the compressor is configured to control attack and release of the compressor by, (i) if the measure of sharpness calculated with application of the deesser weights exceeds one of the thresholds of the compressor, adapting the deesser weights according to a gradient-descent method to attack those parts of the spectral envelope that dominate the measure of sharpness, otherwise, (ii) releasing the deesser weights.