Method and apparatus for suppressing wind noise
View Patent ↗The invention includes a method, apparatus, and computer program to selectively suppress wind noise while preserving narrow-band signals in acoustic data. Sound from one or several microphones is digitized into binary data. A time-frequency transform is applied to the data to produce a series of spectra. The spectra are analyzed to detect the presence of wind noise and narrow band signals. Wind noise is selectively suppressed while preserving the narrow band signals. The narrow band signal is interpolated through the times and frequencies when it is masked by the wind noise. A time series is then synthesized from the signal spectral estimate that can be listened to. This invention overcomes prior art limitations that require more than one microphone and an independent measurement of wind speed. Its application results in good-quality speech from data severely degraded by wind noise.
1. A method for attenuating noise in a signal detected by a sound detector, comprising:
converting the signal detected by the sound detector into a set of digital samples representing a single channel of acoustic data associated with a single microphone;
storing the set of digital samples in a data storage device;
performing a time-frequency transform on the set of digital samples to obtain transformed data;
performing signal analysis on the transformed data, by a hardware processor, to identify wind noise in the transformed data, where the step of performing the signal analysis comprises:
measuring one or more characteristics of the transformed data by the hardware processor by identifying signal segments of the signal that lack a time-varying quasi-periodic amplitude and phase and designating those signal segments as wind noise associated with wind striking a portion of the sound detector; and
discriminating between the wind noise and a signal of interest in the transformed data by comparing the harmonic structure of the signal segments of the signal to the harmonic structure of other signal segments of the signal that have a time varying periodic amplitude and a phase modulated sinusoid characteristic by the hardware processor; and
attenuating at least a portion of the wind noise identified in the transformed data at frequencies dominated by wind noise;
where the discriminating between the wind noise and the signal of interest occurs on the output of the single microphone that sources the single channel of the acoustic data.
2. The method of claim 1 , where the step of performing signal analysis further comprises:
analyzing features of a spectrum of the transformed data;
assigning evidence weights based on the step of analyzing; and
processing the evidence weights to determine whether wind noise is present in the spectrum of the transformed data.
3. The method of claim 1 , where the step of performing signal analysis further comprises identifying peaks in a spectrum of the transformed data that have a Signal to Noise Ratio (SNR) exceeding a peak threshold as peaks not stemming from wind noise.
4. The method of claim 1 , where the step of performing signal analysis further comprises identifying peaks in a spectrum of the transformed data that are sharper and narrower than a selected criteria as peaks stemming from a signal of interest.
5. The method of claim 4 , where the step of identifying comprises measuring peak widths by taking an average difference between a highest point and its neighboring points on each side.
6. The method of claim 1 , where the step of performing signal analysis further comprises:
determining a stability of peaks by comparing peaks in a current spectra of the transformed data to peaks from a previous spectra of the transformed data; and
identifying stable peaks as peaks not stemming from wind noise.
7. The method of claim 1 , where the step of performing signal analysis further comprises:
identifying peaks whose phase and amplitude differences exceed a difference threshold as peaks stemming from wind noise.
8. The method of claim 1 , where the step of performing signal analysis further comprises:
fitting a line to a portion of a spectrum of the transformed data;
comparing a slope of the line to a pre-defined threshold; and
determining whether wind noise is present in the spectrum of the transformed data based on the slope.
9. The method of claim 1 , where the step of performing signal analysis further comprises:
fitting a line to a portion of a spectrum of the transformed data;
comparing an intersection point of the line to a pre-defined threshold; and
determining whether wind noise is present in the spectrum of the transformed data based on the intersection point.
10. An apparatus comprising a single channel of acoustic data from a single microphone, comprising:
a data storage device for storing digital data;
a time-frequency transform component configured to transform signals sourced from a single channel of acoustic data into frequency-based digital data representing the single channel of acoustic data associated with the single microphone;
a signal analyzer configured to identify wind noise in the frequency-based digital data, where the signal analyzer comprises a hardware processor configured to store and measure one or more characteristics of the frequency-based digital data indicative of wind pressure fluctuations associated with wind striking a portion of the single microphone by identifying signal segments of the signal that lack a time-varying quasi-periodic amplitude and phase and discriminate between the wind noise and a signal of interest in the frequency-based digital data by comparing the harmonic structure of the signal segments of the signal to the harmonic structure of other signal segments of the signal that have a time varying periodic amplitude and a phase modulated sinusoid characteristic; and
a wind noise attenuation component configured to attenuate at least a portion of the wind noise in the frequency-based digital data using results obtained from the signal analyzer;
where the signal analyzer discriminates between the wind noise and the signal of interest by processing the output of the single microphone that sources the single channel of the acoustic data.
11. The apparatus of claim 10 , where the signal analyzer is configured to:
analyze features of a spectrum of the frequency-based digital data;
assigning evidence weights based on the step of analyzing; and
processing the evidence weights to determine whether wind noise is present in the spectrum of the frequency-based digital data.
12. The apparatus of claim 10 , where the signal analyzer is configured to identify peaks in a spectrum of the frequency-based digital data that have a Signal to Noise Ratio (SNR) exceeding a peak threshold as peaks not stemming from wind noise.
13. The apparatus of claim 10 , where the signal analyzer is configured to identify peaks in a spectrum of the frequency-based digital data that are sharper and narrower than a selected criteria as peaks stemming from a signal of interest.
14. The apparatus of claim 13 , where the signal analyzer is configured to measure peak widths by taking an average difference between a highest point and its neighboring points on each side.
15. The apparatus of claim 10 , where the signal analyzer is configured to:
determine a stability of peaks by comparing peaks in a current spectra of the frequency-based digital data to peaks from a previous spectra of the frequency-based digital data; and
identify stable peaks as peaks not stemming from wind noise.
16. The apparatus of claim 10 , where the signal analyzer is configured to:
identify peaks whose phase and amplitude differences exceed a difference threshold as peaks stemming from wind noise.
17. The apparatus of claim 10 , where the signal analyzer is configured to:
fit a line to a portion of a spectrum of the frequency-based digital data;
compare a slope of the line to a pre-defined threshold; and
determine whether wind noise is present in the spectrum of the frequency-based digital data based on the slope.
18. The apparatus of claim 10 , where the signal analyzer is configured to:
fit a line to a portion of a spectrum of the frequency-based digital data;
compare an intersection point of the line to a pre-defined threshold; and
determine whether wind noise is present in the spectrum of the frequency-based digital data based on the intersection point.
19. A computer program product, comprising:
a non-transitory computer usable storage medium having computer readable program code embodied therein configured for suppressing noise, comprising:
computer readable code configured to cause a computer to perform a time-frequency transform on the signal to obtain transformed data representing a single channel of acoustic data associated with a single microphone;
computer readable code configured to cause the computer to perform signal analysis on the transformed data to identify wind noise in the transformed data, where the computer readable code configured to cause the computer to perform the signal analysis comprises:
computer readable code configured to cause the computer to measure one or more characteristics of the transformed data indicative of wind pressure fluctuations associated with wind striking a portion of the single microphone by identifying signal segments of the signal that lack a time-varying quasi-periodic amplitude and phase; and
computer readable code configured to cause the computer to discriminate between the wind noise and a signal of interest in the transformed data by comparing the harmonic structure of the signal segments of the signal to the harmonic structure of other signal segments of the signal that have a time varying periodic amplitude and a phase modulated sinusoid characteristic; and
computer readable code configured to cause the computer to attenuate at least a portion of the wind noise identified in the transformed data at frequencies dominated by wind noise;
where the discriminating between the wind noise and the signal of interest occurs on the output of the single microphone that sources the single channel of the acoustic data.
20. The computer program product of claim 19 , where the computer readable code configured to cause the computer to perform signal analysis further comprises:
computer readable code configured to cause the computer to fit a line to a portion of a spectrum of the transformed data;
computer readable code configured to cause the computer to compare a slope of the line and an intersection point of the line to a plurality of pre-defined thresholds; and
computer readable code configured to cause the computer to determine whether wind noise is present in the spectrum of the transformed data based on the slope and the intersection point.