Pitch detection method and apparatus

• US 20050091045A1
• Filed: 10/21/2004
• Published: 04/28/2005
• Est. Priority Date: 10/25/2003
• Status: Active Grant

• Current Assignee: Samsung Electronics Co. Ltd.
• Original Assignee: Samsung Electronics Co. Ltd.
• Inventors: Oh, Kwangcheol
• Granted Patent: US 7,593,847 B2
• Time in Patent Office: Days
• US Class Current: 704/207
• CPC Class Codes: G10L 25/90 Pitch determination of speech signals

Abstract

A pitch detection method and apparatus, the pitch detection apparatus includes: a data rearrangement unit which rearranges voice data on the basis of a center peak of the voice data included in a single frame; a decomposition unit which decomposes rearranged voice data into even symmetrical components on the basis of a center peak; a pitch determination unit which obtains a segment correlation value between a reference point and at least one or more local peaks in relation to even symmetrical components, and determines the location of a local peak corresponding to a maximum segment correlation value among the obtained segment correlation values, as a pitch period.

Claims

1. A pitch detection method comprising:

decomposing voice data into even-number symmetrical components on a basis of a center peak of the voice data included in a single frame; and

determining a location of a local peak corresponding to a maximum segment correlation value among segment correlation values between a reference point and at least one or more local peaks in relation to the even-number symmetrical components, as a pitch period.

2. The pitch detection method of claim 1, wherein the decomposing of the voice data comprises:

multiplying the voice data of the single frame by a first weight window function and then detecting the center peak where an absolute value of a result of the multiplication is a maximum;

shifting the voice data of the single frame on the basis of the center peak; and

decomposing the voice data of the single frame into even symmetrical components on the basis of the center peak.

3. The pitch detection method of claim 1, wherein the decomposing of the voice data comprises:

multiplying the voice data of the single frame by a first weight window function and then detecting the center peak where an absolute value of a result of the multiplication is a maximum;

shifting the voice data of the single frame on the basis of the center peak; and

multiplying the voice data of the single frame by a second weight window function and then decomposing the voice data of the single frame multiplied by the second weight window function, into even symmetrical components on the basis of the center peak.

4. The pitch detection method of claim 2, wherein the first weight window function is any one of Triangular, Hanning, Hamming, Blackmann, Welch or Blackmann-Harris windows functions.

5. The pitch detection method of claim 3, wherein the first weight window function is any one of Triangular, Hanning, Hamming, Blackmann, Welch or Blackmann-Harris windows functions.

6. The pitch detection method of claim 3, wherein the second weight window function is any one of Hanning or Hamming window functions.

7. The pitch detection method of claim 2, further comprising before the decomposing of the voice data:

performing low pass filtering of the voice data being input.

8. The pitch detection method of claim 3, further comprising before the decomposing of the voice data:

performing low pass filtering of the voice data being input.

9. The pitch detection method of claim 1, wherein the determining of the pitch period comprises:

selecting the maximum segment correlation value among obtained segment correlation values;

comparing the maximum segment correlation value with a predetermined threshold; and

if the maximum segment correlation value is greater than the predetermined threshold, determining the location of the local peak corresponding to the maximum segment correlation value, as the pitch period.

10. The pitch detection method of claim 1, wherein the local peak is detected in any one of a negative number area and a positive number area according to a value of the center peak.

11. A computer readable recording medium having embodied thereon a computer program for a pitch detection method comprising:

decomposing voice data into even-number symmetrical components on a basis of a center peak of the voice data included in a single frame; and

determining a location of a local peak corresponding to a maximum segment correlation value among segment correlation values between a reference point and at least one or more local peaks in relation to the even-number symmetrical components, as a pitch period.

12. A pitch detection apparatus comprising:

a decomposition unit which decomposes voice data into even-number symmetrical components on a basis of a center peak of the voice data included in a single frame; and

a pitch determination unit which determines a location of a local peak corresponding to a maximum segment correlation value among segment correlation values between a reference point and at least one or more local peaks in relation to the even-number symmetrical components, as a pitch period.

13. The pitch detection apparatus of claim 12, further comprising a data rearrangement unit which rearranges the voice data on the basis of the center peak of the voice data included in the single frame and provides the rearranged voice data to the decomposition unit.

14. The pitch detection apparatus of claim 13, wherein the data rearrangement unit comprises:

a center peak determination unit which multiplies the voice data of the single frame by a first weight window function and then determines the center peak where an absolute value of the multiplication is a maximum; and

a data transition unit which shifts the voice data of the single frame on the basis of the center peak.

15. The pitch detection apparatus of claim 12, wherein the decomposition unit multiplies the voice data of the single frame by a second weight window function and then decomposes the voice data of the single frame multiplied by the second weight window function, into the even symmetrical components on the basis of the center peak.

16. The pitch detection apparatus of claim 12, wherein the pitch determination unit comprises:

a local peak detection unit which detects at least one or more local peaks in relation to the even symmetrical components;

a correlation value calculation unit which obtains a segment correlation value between the reference point and each of the local peaks; and

a pitch period determination unit which selects the maximum segment correlation value among the obtained segment correlation values, and if the maximum segment correlation value is greater than a predetermined threshold, determines the location of the local peak corresponding to the maximum segment correlation value, as the pitch period.

17. The pitch detection apparatus of claim 12, wherein the local peak is detected in any one of a negative number area and a positive number area according to a value of the center peak.

18. A pitch detection apparatus comprising:

a data rearrangement unit shifting voice data based on a determined center peak included in a single frame unit;

a decomposition unit decomposing the shifted voice data into even-number symmetrical components; and

a pitch determination unit determining a location of a local peak corresponding to a maximum segment correlation value among segment correlation values between a reference point and at least one or more local peaks in relation to the even-number symmetrical components, as a pitch period.

19. The pitch detection apparatus of claim 18, wherein the data rearrangement unit comprises:

a filter unit filtering the voice data;

a frame forming unit dividing the voice data in predetermined time units and forming frame units;

a center peak determination unit multiplying the voice data by a predetermined weight window and determining a location where an absolute value of the multiplication is a maximum, as a center peak; and

a data transition unit shifting the voice data based on the determined center peak so that the center peak is placed at a center of the voice data.

20. The pitch detection apparatus of claim 18, wherein the pitch determination unit comprises:

a local peak detection unit detecting local peaks from the even-number symmetrical components;

a correlation value calculation unit obtaining segment correlation values between a reference point and each of the local peaks detected by the local peak detection unit; and

a pitch period determination unit selecting a maximum segment correlation value among the segment correlation values, and if the maximum segment correlation value is greater than a predetermined threshold, determining the location of the local peak used to obtain the maximum segment correlation value, as a pitch period.

21. The pitch detection apparatus of claim 18, wherein the local peak is detected in any one of a negative number area or a positive number area according to the center peak.

22. A pitch detection method comprising:

shifting voice data based on a determined center peak included in a single frame unit;

decomposing the shifted voice data into even-number symmetrical components; and

determining a location of a local peak corresponding to a maximum segment correlation value among segment correlation values between a reference point and at least one or more local peaks in relation to the even-number symmetrical components, as a pitch period.

23. The pitch detection method of claim 22, wherein the shifting of the voice data further comprises:

filtering the voice data;

dividing the voice data in predetermined time units and forming frame units;

multiplying the voice data by a predetermined weight window and determining a location where an absolute value of the multiplication is a maximum, as a center peak; and

shifting the voice data based on the determined center peak so that the center peak is placed at a center of the voice data.

24. The pitch detection method of claim 22, wherein the determining of the location of the local peak corresponding to the maximum segment correlation value comprises:

detecting local peaks from the even-number symmetrical components;

obtaining segment correlation values between a reference point and each of the detected local peaks; and

selecting a maximum segment correlation value among the segment correlation values, and if the maximum segment correlation value is greater than a predetermined threshold, determining the location of the local peak used to obtain the maximum segment correlation value, as a pitch period.