Method and apparatus for extracting pitch information from audio signal using morphology

• US 7,822,600 B2
• Filed: 07/11/2006
• Issued: 10/26/2010
• Est. Priority Date: 07/11/2005
• Status: Expired due to Fees

• Current Assignee: Samsung Electronics Co. Ltd.
• Original Assignee: Samsung Electronics Co. Ltd.
• Inventors: Kim, Hyun-Soo
• Primary Examiner(s): Jackson; Jakieda R
• Application Number:
  US11/484,204
• Publication Number:
  US 20070106503A1
• Time in Patent Office: 1,568 Days
• Field of Search: 704/207, 704/211
• US Class Current: 704/207
• CPC Class Codes: G10L 25/90 Pitch determination of speech signals

Abstract

A function of improving accuracy of the extraction of pitch information in an audio signal including voice and sound signals is implemented. To do this, a morphological operation is used. In detail, an input audio signal is converted to an audio signal in a frequency domain, an optimum structuring set size (SSS) is determined, and a morphological operation is performed using the determined SSS. Then, by extracting the highest peak from a signal obtained through a predetermined fold and summation process as pitch information, the pitch information can be used in all audio systems in the latter part when voice coding, recognition, synthesis, and/or robustness are performed.

Claims

1. A method of extracting pitch information from an audio signal, comprising the steps of:

when the audio signal is input, converting, by a frequency domain converter, the audio signal to a frequency domain

(b) determining an optimum window size for extracting a pitch from the converted audio signal;

(c) calculating a maximum value and a minimum value of the converted audio signal in optimum window using the determined optimum window size;

(d) checking a variation between the maximum value and the minimum value and generating a staircase signal that has the minimum value in the variation and is used for filtering;

(e) generating a residual signal by extracting the generated staircase signal from the converted audio signal;

(f) generating pitch information by selecting a highest peak generated by performing a predetermined fold and summation process for folding and summing the residual signal; and

(g) extracting the pitch information from the residual signal corresponding to the extraction result,

wherein the staircase signal includes a plurality of flat signals continuously connected, each flat signal having a constant amplitude in a corresponding optimum window for a morphological operation.

2. The method of claim 1, wherein the input audio signal is one of a voice signal and a sound signal.

3. The method of claim 1, wherein step (b) includes searching on a one-by-one basis from pre-set window sizes.

4. The method of claim 3, wherein the searching step includes adjusting the optimum window size based on a number of peaks selected from a pre-processed audio signal.

5. The method of claim 4, wherein the adjusting includes, after defining the number of the selected peaks as N, calculating a ratio of energy of N selected peaks to remaining non-selected peaks by means of the N selected peaks, and determining an optimum window size by comparing the calculated energy ratio and the selected optimum window size.

6. The method of claim 1, wherein the optimum window size is predetermined according to a type of the input audio signal.

7. The method of claim 1, wherein step (c) includes a dilation operation for calculating the maximum value of the audio signal in a predetermined threshold and an erosion operation for calculating the minimum value of the audio signal.

8. The method of claim 7, wherein step (c) includes an opening operation for smoothing by the dilation operation, followed by the erosion operation, and a closing operation for filling by the dilation operation, followed by the erosion operation.

9. The method of claim 1, wherein step (d) includes generating the staircase signal by repeatedly filtering all of the converted audio signals.

10. The method of claim 1, wherein steps (c) and (d) are repeatedly performed on the input audio signal.

11. An apparatus for extracting pitch information from an audio signal, comprising:

a frequency domain converter for converting an input audio signal in a time domain to an audio signal in a frequency domain;

a determiner for determining an optimum window size for extracting a pitch from the converted audio signal;

a calculator for calculating a maximum value and a minimum value of the converted audio signal in an optimum window using the determined optimum window size;

a filter for checking a variation between the maximum value and the minimum value, generating a staircase signal that has the minimum value in the variation, and extracting the generated staircase signal from the converted audio signal; and

an extractor for extracting pitch information from a residual signal corresponding to the extraction result,

wherein the staircase signal includes a plurality of flat signals continuously connected, each flat signal having a constant amplitude in a corresponding optimum window for a morphological operation, the residual signal is a signal obtained by removing the staircase signal from the converted audio signal and the pitch information is a highest peak generated by performing a predetermined fold and summation process for folding and summing the residual signal.

12. The apparatus of claim 11, wherein the input audio signal is one of a voice signal and a sound signal.

13. The apparatus of claim 11, wherein the determiner searches on a one-by-one basis from pre-set window sizes.

14. The apparatus of claim 11, wherein the determiner adjusts the optimum window size based on a number of peaks selected from a pre-processed audio signal.

15. The apparatus of claim 14, wherein the adjusting includes, after defining the number of the selected peaks as N, calculating a ratio of energy of N selected peaks to remaining non-selected peaks by means of the N selected peaks, and determining an optimum window size by comparing the calculated energy ratio and the selected optimum window size.

16. The apparatus of claim 11, wherein the optimum window size is predetermined according to a type of the input audio signal.

17. The apparatus of claim 11, wherein the calculator further performs a dilation operation for calculating the maximum value of the audio signal in a predetermined threshold and an erosion operation for calculating the minimum value of the audio signal.

18. The apparatus of claim 17, wherein the calculator further performs an opening operation for smoothing by the dilation operation, followed by the erosion operation, and a closing operation for filling by the dilation operation, followed by the erosion operation.

19. The apparatus of claim 11, wherein the filter further generates the staircase signal by repeatedly filtering all of the converted audio signals.

20. The apparatus of claim 11, wherein each operation in the calculator and the filter are repeatedly performed on the input audio signal.