Method of using a dominant angle of incidence to reduce acoustic noise in a speech signal
First Claim
1. Method of reducing acoustic noise in a received sound signal incorporating a speech signal by subtracting spectral components of noise from said received sound signal to reconstitute a speech signal spectrum, said method comprising the steps of:
- applying Fourier transform processing to digital signals obtained by analog-to-digital conversion of signals supplied by two microphones spaced by a fixed distance and receiving said sound signal to produce two series of discrete data each discrete datum of which being associated with a respective frequency band of the spectrum of said sound signal and representing the energy magnitude and phase of said respective frequency band, said each of said two series being representative of both noise and speech components of said received sound signal spectrum, each individual difference in phase between the data of the two series associated with the same frequency band representing a respective individual angle of incidence of a respective portion of the received sound signal;
determining a single dominant angle of incidence of said received sound signal from a plurality of the individual angles of incidence, said dominant angle of incidence representing an assumed angle of incidence of said speech signal;
combining said two series to thereby obtain a single instantaneous spectrum of said received sound signal;
for each frequency band of said instantaneous spectrum, comparing a respective tolerance threshold value with a respective absolute value of a respective difference between said dominant angle of incidence and said respective angle of incidence,forming a noise spectrum comprising the same frequency bands as the frequency bands of the instantaneous spectrum,updating the noise spectrum with the energy magnitude data associated with those frequency bands of the instantaneous spectrum for which the respective absolute value is greater than said respective tolerance threshold value to thereby form an updated noise spectrum; and
subtracting said updated noise spectrum from said instantaneous spectrum to obtain an output spectrum approximating said speech signal spectrum.
1 Assignment
0 Petitions
Accused Products
Abstract
Fourier transform processing is applied to digital signals obtained by analog-to-digital conversion of signals supplied by two microphones spaced by a fixed distance to produce two series of discrete data each datum of which represents the energy and phase of a spectral frequency band of the received sound. A dominant angle of incidence representing the angle of incidence of a speech signal component of the received sound signal relative to the two microphones is determined from phase differences between the discrete data in the same frequency bands of the two series and is used to combined the two series of discrete data into a single instantaneous spectrum in which any speech signal component is amplified relative to the noise. A noise spectrum is updated by comparing for each frequency band of the instantaneous spectrum the absolute value of the difference between the dominant angle of incidence and the angle of incidence of the frequency band in question with a tolerance threshold, and setting the energy of the noise spectrum equal to that of the instantaneous spectrum if that difference is above the tolerance threshold. The updated noise spectrum is then subtracted from the instantaneous spectrum to obtain an output spectrum consisting of the speech spectrum.
-
Citations
10 Claims
-
1. Method of reducing acoustic noise in a received sound signal incorporating a speech signal by subtracting spectral components of noise from said received sound signal to reconstitute a speech signal spectrum, said method comprising the steps of:
-
applying Fourier transform processing to digital signals obtained by analog-to-digital conversion of signals supplied by two microphones spaced by a fixed distance and receiving said sound signal to produce two series of discrete data each discrete datum of which being associated with a respective frequency band of the spectrum of said sound signal and representing the energy magnitude and phase of said respective frequency band, said each of said two series being representative of both noise and speech components of said received sound signal spectrum, each individual difference in phase between the data of the two series associated with the same frequency band representing a respective individual angle of incidence of a respective portion of the received sound signal; determining a single dominant angle of incidence of said received sound signal from a plurality of the individual angles of incidence, said dominant angle of incidence representing an assumed angle of incidence of said speech signal; combining said two series to thereby obtain a single instantaneous spectrum of said received sound signal; for each frequency band of said instantaneous spectrum, comparing a respective tolerance threshold value with a respective absolute value of a respective difference between said dominant angle of incidence and said respective angle of incidence, forming a noise spectrum comprising the same frequency bands as the frequency bands of the instantaneous spectrum, updating the noise spectrum with the energy magnitude data associated with those frequency bands of the instantaneous spectrum for which the respective absolute value is greater than said respective tolerance threshold value to thereby form an updated noise spectrum; and subtracting said updated noise spectrum from said instantaneous spectrum to obtain an output spectrum approximating said speech signal spectrum. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9)
-
-
10. Method of reducing acoustic noise in a received sound signal comprising a speech signal and a noise signal comprising the steps of:
-
determining a respective sound signal spectrum from each of two sources, each said spectrum including a plurality of adjacent frequency bands and a respective energy for each frequency band; calculating a respective angle of incidence for each frequency band representative of a respective phase difference of the two sources within the respective frequency band; calculating an individual weighting for each frequency band proportional to the energy and frequency of the two signal spectra within the respective frequency band; establishing a plurality of disjoint ranges of angle of incidence; for each range of said disjoint ranges, calculating a sliding mean of a sum of the individual weightings of those frequency bands whose angle of incidence is within the respective range, to thereby establish a respective cumulative weighting; identifying a greatest cumulative weighting; selecting a dominant angle of incidence within the respective range corresponding to said greatest cumulative weighting; setting a relative phase between the two spectra equal to said dominant angle of incidence to thereby obtain two phase adjusted series of discrete data; adding the respective discrete data of the two phase adjusted series to obtain an amplified instantaneous spectrum incorporating both said sound signal and said noise signal; calculating for each frequency band of said instantaneous spectrum a respective absolute value of a difference between said dominant angle of incidence and the respective angle of incidence of the frequency band; comparing each said absolute value with a tolerance threshold value; obtaining an updated noise spectrum from the instantaneous spectrum, each frequency band of said updated noise spectrum having an energy equal to a corresponding value of the instantaneous spectrum if the respective absolute value is greater than said tolerance threshold value; subtracting said updated noise spectrum from said instantaneous spectrum to obtain an output spectrum comprising a corresponding spectrum of said speech signal; counting, following said subtraction, the number of frequency bands in which the energy of the output spectrum is above an energy threshold value; if said number is greater than a predetermined value indicative of said output spectrum containing a large number of meaningful data and being essentially representative of said speech signal, setting the energy of each band of a corrected noise spectrum to the lessor of the energy of the respective frequency band of said instantaneous spectrum or the energy of said updated noise spectrum; and if said number is not greater than said predetermined value and said output spectrum does not contain said large number of meaningful data and said instantaneous spectrum is therefore essentially representative of said noise signal, setting to zero the energy of those frequency bands of said output spectrum for which the respective energy is above said energy threshold value and replacing the energy of at least the corresponding frequency bands of said corrected noise spectrum with the respective energies of the instantaneous spectrum.
-
Specification