# US 12,763,091 B2 · Enhanced de-esser for in-car communications systems > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 12,763,091 B2 - Document ID: US12763091B2 - Status: Granted - Kind: B2 - Application number: 18/386,825 - Filed: 2023-11-03 - Published: 2024-02-22 - Granted: 2026-10-13 - Earliest claimed: 2016-05-11 - Assignees: Cerence Operating Company - Applicants: Cerence Operating Company - Inventors: Tobias Herbig; Stefan Richardt - Web page: https://patentgrep.com/patent/US12763091B2 ## CPC - G10L 21/0364: for improving intelligibility - G10L 25/18: the extracted parameters being spectral information of each sub-band - H03G 9/025: {frequency-dependent volume compression or expansion, e.g. multiple-band systems (H03G9/10, H03G9/18 take precedence)} - G10L: SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING - H03G: CONTROL OF AMPLIFICATION ## Abstract 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. ## Claims ### Claim 1 A method comprising, in an in-car communication system, de-essing a speech signal, wherein de-essing said speech signal comprises, for each time frame of a sequence of time frames: receiving, at a speech-processing system, a full spectral envelope that comprises a combination of said speech signal and background noise, said full spectral envelope consisting of a first part and a second part; analyzing said full spectral envelope to identify frequency content for de-essing; and spectrally weighting said speech signal to carry out said de-essing, wherein spectrally weighting said speech signal is based on both said first part and said second part, wherein said second part is devoid of sibilant sounds, the method further comprising, after having spectrally weighted said speech signal to carry out said de-essing, determine a first measure of sharpness and a second measure of sharpness, said first measure being based on a result of having carried out said de-essing and said second measure being based on a result of not having carried out said de-essing, determining that said first measure fails to exceed a threshold, and stopping said de-essing of said speech signal. ### Claim 2 (depends on 1) The method of claim 1, wherein spectrally weighting said speech signal to carry out said de-essing based on both said first part and said second part comprises determining weights based on a psychoacoustic measure that has been obtained from said full spectral envelope. ### Claim 3 (depends on 1) The method of claim 1, wherein spectrally weighting said speech signal to carry out said de-essing based on both said first part and said second part comprises determining weights based on sibilant sounds that have been detected as a result of using a psychoacoustic measure that has been obtained from said first and second parts. ### Claim 4 (depends on 1) The method of claim 1, wherein spectrally weighting said speech signal to carry out said de-essing based on both said first part and said second part comprises determining weights based on a measure of sharpness that has been obtained based on a combination of said first part of said full spectral envelope and said second part of said full spectral envelope. ### Claim 5 (depends on 1) The method of claim 1, wherein spectrally weighting said speech signal to carry out said de-essing based on both said first part and said second part comprises determining weights based on a measure of roughness that has been obtained based on said first and second parts. ### Claim 6 (depends on 1) The method of claim 1, wherein said speech signal comprises sibilant sounds and wherein spectrally weighting said speech signal to carry out said de-essing based on both said first part and said second part comprises applying de-esser weights to said sibilant sounds of said speech signal. ### Claim 7 (depends on 1) The method of claim 1, wherein spectrally weighting said speech signal to carry out said de-essing comprising determining weights to be used for said de-essing and wherein said method further comprises applying said weights to control attack and release of a compressor. ### Claim 8 (depends on 1) The method of claim 1, wherein spectrally weighting said speech signal to carry out said de-essing comprises spectrally weighting at a frequency resolution that matches that of said full spectral envelope. ### Claim 9 A method executed by an in-car communication system for processing a speech signal, said method comprising: receiving a first signal; for each frequency component in said first signal, multiplying said frequency component by a corresponding weight, said weights being updated over time to reduce a first frequency-independent psychoacoustic parameter that has a value that changes in response to changes in said first signal; after having received said first signal, receiving a second signal; and, for each frequency component in said second signal, multiplying said frequency component by a corresponding weight, said weights being updated over time in a manner that is independent of a second frequency-independent psychoacoustic parameter that has a value that changes in response to changes in said second signal, wherein said method further comprises an attack phase, in which said first signal is received, and a release phase, in which said second signal is received, wherein said method further comprises causing a transition between said attack phase and said release phase based on an output of a voice-activity detector. ### Claim 10 (depends on 9) The method of claim 9, further comprising selecting said psychoacoustic parameter to be roughness. ### Claim 11 (depends on 9) The method of claim 9, wherein receiving said first signal comprises receiving a first signal at a compressor and wherein said compressor is configured to update said weights to reduce said first frequency-independent psychoacoustic parameter. ### Claim 12 (depends on 9) The method of claim 9, further, while receiving said first signal, comprising updating said weights by decrementing a corresponding one of said weights by a value that is proportional to said first frequency-independent parameter. ### Claim 13 (depends on 9) The method of claim 9, further comprising, while receiving said first signal, updating said weights by decrementing a corresponding one of said weights by a value that is proportional to a ratio between said first frequency-independent parameter and a difference between a pair of preceding weights. ### Claim 14 (depends on 9) The method of claim 9, further comprising detecting, in said speech signal, a transition between said first signal and said second signal. ### Claim 15 (depends on 9) The method of claim 9, further comprising detecting onset of voice activity in said speech signal and, as a result thereof, causing said speech signal to be processed as said first signal. ### Claim 16 (depends on 9) The method of claim 9, further comprising detecting cessation of voice activity in said speech signal and, as a result thereof, causing said speech signal to be processed as said second signal. ### Claim 17 (depends on 9) The method of claim 9, further comprising re-setting said weights in response to determining that no attenuation of fricatives in said speech signal is needed. ### Claim 18 (depends on 9) The method of claim 9, further comprising selecting said psychoacoustic parameter to be sharpness.