Abstract: Provided is a noise elimination device capable of responding to a change in noise over time using a single acoustic sensor. The present invention comprises a Fourier transform unit for dividing a single input signal including a component other than a target signal into frames delimited by a fixed time to convert the input signal to a frequency spectrum, and using a delay amount to generate a plurality of pairs of frequency spectra having a fixed time difference, a cross spectrum unit for generating a plurality of cross spectra or coherences from the plurality of pairs of frequency spectra having a fixed time difference, and a noise elimination unit for averaging the plurality of cross spectra or plurality of coherences and extracting a power spectrum of the target signal.
We Claim:
[Claim 1] A noise removal device comprising:
a Fourier transform unit to divide one input signal, the input signal containing
a component other than a target signal, into frames sectioned by fixed time and transform the input signal into a frequency spectrum so as to generate a plurality of pairs of frequency spectra, the pairs of frequency spectra having a fixed time difference, by using a delay amount;
a cross spectrum unit to generate a plurality of cross spectra or a plurality of coherences from the plurality of pairs of frequency spectra having the fixed time difference; and
a noise removal unit to average the plurality of cross spectra or the plurality of coherences so as to extract a power spectrum of the target signal.
[Claim 2] The noise removal device according to claim 1, wherein
the Fourier transform unit transforms the input signal into a frequency spectrum for each frame and combines a plurality of said frequency spectra so as to generate a first array of the frequency spectra,
the Fourier transform unit further includes a delay unit,
the delay unit delays the first array of the frequency spectra by using the delay amount so as to generate a second array of the frequency spectra, and
the pairs of frequency spectra having the fixed time difference are pairs composed of elements of the first array of the frequency spectra and elements of the second array of the frequency spectra.
[Claim 3] The noise removal device according to claim 1, wherein
the Fourier transform unit further includes a delay unit,
the delay unit generates a signal obtained by delaying the input signal by using the delay amount,
the Fourier transform unit
transforms the input signal into a frequency spectrum for each frame and combines a plurality of said frequency spectra so as to generate a first array of the frequency spectra, and
transforms the delayed signal into a frequency spectrum for each frame so as to generate a second array of the frequency spectra, the second array of the frequency spectra being delayed compared to the first array of the frequency spectra, and
the pairs of frequency spectra having the fixed time difference are pairs composed of elements of the first array of the frequency spectra and elements of the second array of the frequency spectra.
[Claim 4] The noise removal device according to claim 2 or 3, comprising:
a delay amount selection unit to use the first array of the frequency spectra and the power spectrum of the target signal so as to select the delay amount from a plurality of candidates for a delay amount so that an evaluation value related to noise removal performance is high, wherein
the Fourier transform unit generates the second array of the frequency spectra, the second array of the frequency spectra being delayed compared to the first array of the frequency spectra, by using the delay amount selected.
[Claim 5] The noise removal device according to claim 2 or 3, comprising:
a delay amount calculation unit to use period information related to the target signal so as to calculate the delay amount corresponding to temporal variation of the period information, wherein
the Fourier transform unit generates the second array of the frequency spectra, the second array of the frequency spectra being delayed compared to the first array of the frequency spectra, by using the delay amount calculated.
[Claim 6] The noise removal device according to claim 5, wherein
the delay amount calculation unit calculates a large value of the delay amount when the temporal variation of the period information is small, and calculates a small value of the delay amount when the temporal variation of the period information is large.
[Claim 7] The noise removal device according to any one of claims 1 to 6,
wherein
the delay amount has a different value for each frequency.
[Claim 8] A noise removal method comprising:
by a Fourier transform unit, dividing one input signal, the input signal containing a component other than a target signal, into frames sectioned by fixed time and transforming the input signal into a frequency spectrum so as to generate a plurality of pairs of frequency spectra, the pairs of frequency spectra having a fixed time difference, by using a delay amount;
generating, by a cross spectrum unit, a plurality of cross spectra or a plurality of coherences from the plurality of pairs of frequency spectra having the fixed time difference; and
averaging, by a noise removal unit, the plurality of cross spectra or the plurality of coherences so as to extract a power spectrum of the target signal. [Claim 9] The noise removal method according to claim 8, wherein
the Fourier transform unit transforms the input signal into a frequency spectrum for each frame and combines a plurality of said frequency spectra so as to generate a first array of the frequency spectra, the Fourier transform unit further includes a delay unit,
the delay unit delays the first array of the frequency spectra by using the delay amount so as to generate a second array of the frequency spectra, and
the pairs of frequency spectra having the fixed time difference are pairs composed of elements of the first array of the frequency spectra and elements of the second array of the frequency spectra.
[Claim 10] The noise removal method according to claim 8, wherein
the Fourier transform unit further includes a delay unit,
the delay unit generates a signal obtained by delaying the input signal by using the delay amount,
the Fourier transform unit
transforms the input signal into a frequency spectrum for each frame and combines a plurality of said frequency spectra so as to generate a first array of the frequency spectra, and
transforms the delayed signal into a frequency spectrum for each frame so as to generate a second array of the frequency spectra, the second array of the frequency spectra being delayed compared to the first array of the frequency spectra, and
the pairs of frequency spectra having the fixed time difference are pairs composed of elements of the first array of the frequency spectra and elements of the second array of the frequency spectra.
[Claim 11] The noise removal method according to claim 9 or 10, wherein
a delay amount selection unit uses the first array of the frequency spectra and the power spectrum of the target signal so as to select the delay amount from a plurality of candidates for a delay amount so that an evaluation value related to noise removal performance is high, and
the Fourier transform unit generates the second array of the frequency spectra, the second array of the frequency spectra being delayed compared to the first array of the frequency spectra, by using the delay amount selected.
[Claim 12] The noise removal method according to claim 9 or 10, wherein
a delay amount calculation unit uses period information related to the target signal so as to calculate the delay amount corresponding to temporal variation of the period information, and
the Fourier transform unit generates the second array of the frequency spectra, the second array of the frequency spectra being delayed compared to the first array of the frequency spectra, by using the delay amount calculated.
[Claim 13] The noise removal method according to claim 12, wherein
the delay amount calculation unit calculates a large value of the delay amount when the temporal variation of the period information is small, and calculates a small value of the delay amount when the temporal variation of the period information is large.
[Claim 14] The noise removal method according to any one of claims 8 to 13,
wherein
the delay amount has a different value for each frequency.
| # | Name | Date |
|---|---|---|
| 1 | 202447064926-TRANSLATIOIN OF PRIOIRTY DOCUMENTS ETC. [28-08-2024(online)].pdf | 2024-08-28 |
| 2 | 202447064926-STATEMENT OF UNDERTAKING (FORM 3) [28-08-2024(online)].pdf | 2024-08-28 |
| 3 | 202447064926-REQUEST FOR EXAMINATION (FORM-18) [28-08-2024(online)].pdf | 2024-08-28 |
| 4 | 202447064926-PROOF OF RIGHT [28-08-2024(online)].pdf | 2024-08-28 |
| 5 | 202447064926-PRIORITY DOCUMENTS [28-08-2024(online)].pdf | 2024-08-28 |
| 6 | 202447064926-POWER OF AUTHORITY [28-08-2024(online)].pdf | 2024-08-28 |
| 7 | 202447064926-FORM 18 [28-08-2024(online)].pdf | 2024-08-28 |
| 8 | 202447064926-FORM 1 [28-08-2024(online)].pdf | 2024-08-28 |
| 9 | 202447064926-DRAWINGS [28-08-2024(online)].pdf | 2024-08-28 |
| 10 | 202447064926-DECLARATION OF INVENTORSHIP (FORM 5) [28-08-2024(online)].pdf | 2024-08-28 |
| 11 | 202447064926-COMPLETE SPECIFICATION [28-08-2024(online)].pdf | 2024-08-28 |
| 12 | 202447064926-MARKED COPIES OF AMENDEMENTS [16-09-2024(online)].pdf | 2024-09-16 |
| 13 | 202447064926-FORM 13 [16-09-2024(online)].pdf | 2024-09-16 |
| 14 | 202447064926-AMMENDED DOCUMENTS [16-09-2024(online)].pdf | 2024-09-16 |
| 15 | 202447064926-FORM 3 [30-01-2025(online)].pdf | 2025-01-30 |