# US 10,504,501 B2 · Adaptive suppression for removing nuisance audio > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 10,504,501 B2 - Document ID: US10504501B2 - Status: Granted - Kind: B2 - Application number: 16/073,390 - Filed: 2017-02-02 - Published: 2019-02-28 - Granted: 2019-12-10 - Earliest claimed: 2016-02-02 - Assignees: Dolby Laboratories Licensing Corporation - Applicants: Dolby Laboratories Licensing Corporation - Inventors: Glenn N. Dickins; Dong Shi; David Gunawan - Web page: https://patentgrep.com/patent/US10504501B2 ## CPC - G10K 11/17823: {Reference signals, e.g. ambient acoustic environment} - G10K 11/17853: {of the filter} - H04M 9/08: Two-way loud-speaking telephone systems with means for conditioning the signal, e.g. for suppressing echoes for one or both directions of traffic - G10K 2210/1081: Earphones, e.g. for telephones, ear protectors or headsets - G10K 2210/129: Vibration, e.g. instead of, or in addition to, acoustic noise - G10K 2210/3028: Filtering, e.g. Kalman filters or special analogue or digital filters - G10K: SOUND-PRODUCING DEVICES; METHODS OR DEVICES FOR PROTECTING AGAINST, OR FOR DAMPING, NOISE OR OTHER ACOUSTIC WAVES IN GENERAL; ACOUSTICS NOT OTHERWISE PROVIDED FOR - H04M: TELEPHONIC COMMUNICATION (circuits for controlling other apparatus via a telephone cable and not involving telephone switching apparatus G08) ## Abstract Described herein is a method and audio capture system for suppressing noise due to mechanical disturbances in an audio capture system. In one embodiment a method includes the steps of: a) receiving an input audio signal from a microphone; b) receiving an input reference signal separate from the input audio signal; c) processing the input reference signal to generate a control signal for identifying a noise event; d) in response to the control signal, estimating a contribution of the noise event to the input audio signal by applying an adaptive filter to a sliding temporal window of the control signal in the frequency domain; and e) based on the estimated contribution, selectively modifying the input audio signal to generate an output audio signal in which the noise arising from the noise event is at least partially suppressed. ## Claims ### Claim 1 A method of suppressing noise due to mechanical disturbances in an audio capture system, the method including the steps of: a. receiving an input audio signal from a microphone and transforming the input audio signal into frequency domain; b. receiving an input reference signal from a sensor other than the microphone, wherein the input reference signal is separate from the input audio signal; c. processing the input reference signal to generate a control signal for identifying a noise event, wherein processing the input reference signal includes transforming the input reference signal into the frequency domain; d. in response to the control signal identifying a noise event, estimating a contribution of the noise event to the input audio signal; and e. based on the estimated contribution of the noise event, selectively modifying the input audio signal to generate an output audio signal in which the noise arising from the noise event is at least partially suppressed, wherein transforming the reference signal includes iteratively transforming the reference signal over predetermined time windows to calculate respective temporally evolving reference power spectra indicative of the reference signal, each reference power spectrum including reference power values separated into a plurality of frequency bins; wherein transforming the input audio signal includes iteratively transforming the input audio signal into the frequency domain over the predetermined time windows to calculate respective temporally evolving signal power spectra indicative of the input audio signal, each signal power spectrum including signal power values separated into a like plurality of frequency bins to the reference power spectra; and wherein estimating the contribution of the noise event to the input audio signal involves applying an adaptive filter in the frequency domain to the reference power spectra and determining parameters of the adaptive filter by comparing the signal power spectra and the reference power spectra over a predetermined period of time, wherein the parameters of the adaptive filter are based on a difference between the reference power spectra after applying the adaptive filter and the signal power spectra. ### Claim 2 (depends on 1) A method according to claim 1 including the step of: b2. analyzing the input reference signal to classify potential noise events likely to be associated with an external disturbance; and wherein the processing in step c is based on the classifying in step b2. ### Claim 3 (depends on 2) A method according to claim 2 wherein the step of classifying potential noise events includes specifying a noise event when the reference signal power is greater than a predetermined threshold. ### Claim 4 (depends on 2) A method according to claim 2 wherein step of classifying potential noise events includes specifying a noise event when the reference signal contains predetermined spectral characteristics. ### Claim 5 (depends on 2) A method according to claim 2 wherein, in the absence of detecting a potential noise event, the control signal is the input reference signal. ### Claim 6 (depends on 2) A method according to claim 2 wherein, in the absence of detecting a potential noise event, the control signal is zero. ### Claim 7 (depends on 1) A method according to claim 1 wherein the step of processing the input reference signal includes filtering the input reference signal to generate the control signal. ### Claim 8 (depends on 7) A method according to claim 7 wherein the filtering includes performing signal gating on the reference signal to generate the control signal. ### Claim 9 (depends on 8) A method according to claim 8 wherein the signal gating is triggered by detection of a threshold amplitude of the reference signal. ### Claim 10 (depends on 1) A method according to claim 1 wherein the step of processing the input reference signal includes extending spectral content of the reference signal in spectral domain to generate the control signal. ### Claim 11 (depends on 1) A method according to claim 1 wherein the step of processing the input reference signal includes adjusting dynamic range of the input reference signal to generate the control signal. ### Claim 12 (depends on 1) A method according to claim 1 wherein the step of processing the input reference signal includes modulating the input reference signal with an external signal to generate the control signal. ### Claim 13 (depends on 1) A method according to claim 1 wherein the step of processing the input reference signal includes generating a synthetic control signal having characteristics based on the input reference signal. ### Claim 14 (depends on 1) A method according to claim 1, wherein the predetermined period of time includes time windows during a detected noise event or shortly before and/or after a detected noise event. ### Claim 15 (depends on 1) A method according to claim 1 wherein the predetermined period of time includes a period covering 5 to 15 predetermined time windows. ### Claim 16 (depends on 1) A method according to claim 1 wherein comparing the power spectra includes iteratively performing a spectral subtraction of the reference power spectra from the signal power spectra across different predetermined time windows to determine a relationship between fluctuations of the respective power spectra. ### Claim 17 (depends on 1) A method according to claim 1 wherein comparing the power spectra includes performing a least mean squares method on the respective reference power spectra and signal power spectra over the predetermined period of time. ### Claim 18 (depends on 1) A method according to claim 1 including the step of calculating estimated noise event power contribution values for each frequency bin in each spectrum within the predetermined period of time. ### Claim 19 (depends on 18) A method according to claim 18 wherein the step of selectively modifying the audio signal includes calculating suppression gain coefficients for each frequency bin within the predetermined period of time based on a comparison of the estimated noise event power contribution values with the signal power values within the predetermined period of time. ### Claim 20 (depends on 19) A method according to claim 19 including the step of multiplying the signal power values by the suppression gain coefficients to generate modified signal power values. ### Claim 21 (depends on 20) A method according to claim 20 including the step of inverse transforming the modified signal power values into the time domain to produce a modified audio signal. ### Claim 22 (depends on 1) A method according to claim 1 wherein the input reference signal is a vibration signal from a vibration sensor. ### Claim 23 (depends on 1) A method according to claim 1 wherein the input reference signal is a detection signal from a button press on a user interface. ### Claim 24 (depends on 1) A method according to claim 1 wherein the input reference signal originates from a second microphone. ### Claim 25 (depends on 24) A method according to claim 24 wherein the input reference signal is a filtered version of an audio signal from the second microphone. ### Claim 26 (depends on 1) A method according to claim 1 wherein the input audio signal and input reference signal are incoherent. ### Claim 27 An audio capture system for suppressing noise due to mechanical disturbances, the system including: a first input port for receiving an input audio signal from a microphone; a second input port for receiving an input reference signal from a sensor other than the microphone, wherein the input reference signal is separate from the input audio signal; and a processor configured for: transforming the input audio signal into the frequency domain; processing the input reference signal to generate a control signal for identifying a noise event; in response to the control signal identifying a noise event, estimating a contribution of the noise event to the input audio signal; based on the estimated contribution of the noise event, selectively modifying the input audio signal to generate an output audio signal in which the noise arising from the noise event is at least partially suppressed, wherein processing the input reference signal includes transforming the input reference signal into the frequency domain; wherein transforming the reference signal includes iteratively transforming the reference signal over predetermined time windows to calculate respective temporally evolving reference power spectra indicative of the reference signal, each reference power spectrum including reference power values separated into a plurality of frequency bins; wherein transforming the input audio signal includes iteratively transforming the input audio signal into the frequency domain over the predetermined time windows to calculate respective temporally evolving signal power spectra indicative of the input audio signal, each signal power spectrum including signal power values separated into a like plurality of frequency bins to the reference power spectra; and wherein estimating the contribution of the noise event to the input audio signal involves applying an adaptive filter in the frequency domain to the reference power spectra and determining parameters of the adaptive filter by comparing the signal power spectra and the reference power spectra over a predetermined period of time, wherein the parameters of the adaptive filter are based on a difference between the reference power spectra after applying the adaptive filter and the signal power spectra. ### Claim 28 (depends on 27) An audio capture system according to claim 27 wherein the processor is further configured for: analyzing the input reference signal to classify potential noise events likely to be associated with an external disturbance; and wherein the processing of the input reference signal is based on the classifying.