Very short pitch detection and coding

• US 10,482,892 B2
• Filed: 07/28/2017
• Issued: 11/19/2019
• Est. Priority Date: 12/21/2011
• Status: Active Grant

• Current Assignee: Huawei Technologies Co., Ltd. (Huawei Investment & Holding Co., Ltd.)
• Original Assignee: Huawei Technologies Co., Ltd. (Huawei Investment & Holding Co., Ltd.)
• Inventors: Gao, Yang, Qi, Fengyan
• Primary Examiner(s): Abebe, Daniel
• Application Number:
  US15/662,302
• Publication Number:
  US 20170323652A1
• Time in Patent Office: 844 Days
• Field of Search: 704207
• CPC Class Codes: G10L 19/00 Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis in musical instruments G10H
• CPC Class Codes: G10L 19/09 Long term prediction, i.e. removing periodical redundancies, e.g. by using adaptive codebook or pitch predictor
• CPC Class Codes: G10L 21/003 Changing voice quality, e.g. pitch or formants
• CPC Class Codes: G10L 25/06 the extracted parameters being correlation coefficients
• CPC Class Codes: G10L 25/21 the extracted parameters being power information
• CPC Class Codes: G10L 25/90 Pitch determination of speech signals

Abstract

System and method embodiments are provided for very short pitch detection and coding for speech or audio signals. The system and method include detecting whether there is a very short pitch lag in a speech or audio signal that is shorter than a conventional minimum pitch limitation using a combination of time domain and frequency domain pitch detection techniques. The pitch detection techniques include using pitch correlations in time domain and detecting a lack of low frequency energy in the speech or audio signal in frequency domain. The detected very short pitch lag is coded using a pitch range from a predetermined minimum very short pitch limitation that is smaller than the conventional minimum pitch limitation.

Claims

1. A method for pitch detection, implemented by an encoder, comprising:

determining a value of an initial pitch lag candidate of a current frame of a signal in a range from a second minimum pitch limitation to a first minimum pitch limitation using a time domain pitch detection technique, wherein the first minimum pitch limitation is a pitch limitation value defined in the Code Excited Linear Prediction Technique (CELP) algorithm, and the second minimum pitch limitation is a value smaller than the first minimum pitch limitation, and wherein the signal is a speech signal or an audio signal;

determining whether a lack of low frequency energy in the current frame is detected; and

determining the initial pitch lag candidate is a final pitch lag when the lack of low frequency energy in the current frame is detected.

2. The method of claim 1, wherein determining whether a lack of low frequency energy in the current frame is detected comprising:

determining a first maximum energy of the current frame in a first frequency region from zero to a predetermined minimum frequency, and a second maximum energy of the current frame in a second frequency region from the predetermined minimum frequency to a predetermined maximum frequency;

calculating an energy ratio of the current frame between the first maximum energy and the second maximum energy;

adjusting the energy ratio using an average normalized pitch correlation of the current frame to obtain an adjusted energy ratio;

calculating a smoothed energy ratio of the current frame using the adjusted energy ratio; and

determining a lack of low frequency energy of the current frame is detected when the smoothed energy ratio is greater than a first threshold or the adjusted energy ratio is greater than a second threshold.

3. The method of claim 2, wherein calculating the energy ratio between the first maximum energy and the second maximum energy comprises:

calculating the energy ratio as;

Ratio=Energy1−

Energy0,

where Ratio is the energy ratio, Energy0 is the first maximum energy in decibel (dB) in a first frequency region [0, F_MIN] Hertz (Hz), Energy1 is the second maximum energy in dB in a second frequency region [F_MIN, 900] Hz, F_MIN is the predetermined minimum frequency, and 900 Hz is the predetermined maximum frequency.

4. The method of claim 2, wherein adjusting the energy ratio to obtain the adjusted energy ratio comprises:

adjusting the energy ratio using the average normalized pitch correlation to obtain the adjusted energy ratio as

Ratio⇐

Ratio·

Voicing,

where Voicing is the average normalized pitch correlation;

Ratio on the right side of the equation is the energy ratio before being adjusted; and

Ratio on the left side of the equation is the adjusted energy ratio.

5. The method of claim 4, wherein calculating the smoothed energy ratio comprises:

calculating the smoothed energy ratio according to the adjusted energy ratio as;

LF_EnergyRatio_sm⇐

(15·

LF_EnergyRatio_sm+Ratio)/16,

where LF_EnergyRatio_sm on the left side of the equation is the smoothed energy ratio of the current frame;

LF_EnergyRatio_sm on the right side of the equation is the smoothed energy ratio of a previous frame; and

Ratio is the adjusted energy ratio.

6. The method of claim 2, wherein the average normalized pitch correlation is obtained by:

calculating the average normalized pitch correlation as

Voicing=[R₁(P₁)+R₂(P₂)+R₃(P₃)+R₄(P₄)]/4,

where Voicing is the average normalized pitch correlation, R₁(P₁), R₂(P₂), R₃(P₃), and R₄(P₄) are four normalized pitch correlations calculated for four respective subframes of the current frame, and P₁, P₂, P₃, and P₄ are four pitch candidates, found in a pitch range from PIT_MIN to PIT_MAX, for the four respective subframes, wherein PIT_MIN is the first minimum pitch limitation, and PIT_MAX is a pitch limitation greater than the first minimum pitch limitation.

7. The method of claim 6, wherein each normalized pitch correlation is calculated according to:

8. The method of claim 6, wherein determining the value of the initial pitch lag candidate comprises:

determining the value of the initial pitch lag candidate as;

R(Pitch_Tp)=MAX{R(P),P=PIT_MIN0, . . . ,PIT_MIN}

where R(P) is a normalized pitch correlation for a pitch lag P, Pitch_Tp is the value of the initial pitch lag candidate;

PIT_MIN0 is the second minimum pitch limitation; and

PIT_MIN is the first minimum pitch limitation.

9. The method of claim 8, wherein the normalized pitch correlation is calculated according to:

10. The method of claim 1, wherein determining the initial pitch lag candidate is a final pitch lag when the lack of low frequency energy in the current frame is detected comprises:

determining the initial pitch lag candidate is the final pitch lag when the lack of low frequency energy in the current frame is detected and a smoothed pitch correlation of the initial pitch lag candidate of the current frame is greater than a third threshold.

11. The method of claim 10, wherein the smooth pitch correlation is calculated according to:

Voicing0_sm⇐

(3·

Voicing0_sm+Voicing0)/4

where Voicing0_sm on the left side is the smoothed pitch correlation of the initial pitch lag candidate of the current frame;

Voicing0_sm on the right side is a smoothed pitch correlation of the initial pitch lag candidate of a previous frame; and

Voicing0 is equal to a normalized pitch correlation of the initial pitch lag candidate.

12. The method of claim 6, wherein the first threshold is 35, and the second threshold is 50.

13. The method of claim 10, wherein determining the initial pitch lag candidate is a final pitch lag when the lack of low frequency energy in the current frame is detected comprises:

determining the initial pitch lag candidate is the final pitch lag when the lack of low frequency energy in the current frame is detected, the smoothed pitch correlation of the initial pitch lag candidate of the current frame is greater than the third threshold, and the smoothed pitch correlation of the initial pitch lag candidate of the current frame is greater than a value of a fourth threshold multiplied by a smoothed pitch correlation Voicing_sm of the current frame.

14. The method of claim 13, wherein the smoothed pitch correlation Voicing_sm is calculated according to:

Voicing_sm⇐

(3·

Voicing_sm+Voicing)/4

where Voicing_sm on the left side is the smoothed pitch correlation of the current frame;

Voicing_sm on the right side is a smoothed pitch correlation of a previous frame; and

Voicing is the average normalized pitch correlation.

15. The method of claim 13, wherein the fourth threshold is 0.7.

16. The method of claim 1, wherein the first minimum pitch limitation is 34 and the second minimum pitch limitation is 17 for 12.8 kilohertz (kHz) sampling frequency.

17. The method of claim 1, further comprising:

encoding the final pitch lag.

18. An audio signal encoder, comprising:

a memory storing program instructions, and one or more processors coupled to the memory;

wherein the one or more processors, by executing the program instructions, are configured to;

determine a value of an initial pitch lag candidate of a current frame of a signal in a range from a second minimum pitch limitation to a first minimum pitch limitation using a time domain pitch detection technique, wherein the first minimum pitch limitation is a pitch limitation value defined in the Code Excited Linear Prediction Technique (CELP) algorithm, and the second minimum pitch limitation is a value smaller than the first minimum pitch limitation, and wherein the signal is a speech signal or an audio signal;

determine whether a lack of low frequency energy of in current frame is detected; and

determine the initial pitch lag candidate is a final pitch lag when the lack of low frequency energy in the current frame is detected.

19. The encoder of claim 18, wherein, to determine whether a lack of low frequency energy in the current frame is detected, the one or more processors are configured to:

determine a first maximum energy of the current frame in a first frequency region from zero to a predetermined minimum frequency, and a second maximum energy of the current frame in a second frequency region from the predetermined minimum frequency to a predetermined maximum frequency;

calculate an energy ratio of the current frame between the first maximum energy and the second maximum energy;

adjust the energy ratio using an average normalized pitch correlation of the current frame to obtain an adjusted energy ratio;

calculate a smoothed energy ratio of the current frame using the adjusted energy ratio; and

determine a lack of low frequency energy of the current frame is detected when the smoothed energy ratio is greater than a first threshold or the adjusted energy ratio is greater than a second threshold.

20. The encoder of claim 19, wherein, to calculate the energy ratio between the first maximum energy and the second maximum energy, the one or more processors are configured to:

calculate the energy ratio as;

Ratio=Energy1−

Energy0,

where Ratio is the energy ratio, Energy0 is the first maximum energy in decibel (dB) in a first frequency region [0, F_MIN] Hertz (Hz), Energy1 is the second maximum energy in dB in a second frequency region [F_MIN, 900] Hz, and F_MIN is the predetermined minimum frequency, and 900 Hz is the predetermined maximum frequency.

21. The encoder of claim 19, wherein, to adjust the energy ratio to obtain the adjusted energy ratio, the one or more processors are configured to:

adjust the energy ratio using the average normalized pitch correlation to obtain the adjusted energy ratio of the current frame as

Ratio⇐

Ratio·

Voicing,

where Voicing is the average normalized pitch correlation;

Ratio on the right side of the equation is the energy ratio before being adjusted; and

Ratio on the left side of the equation is the adjusted energy ratio.

22. The encoder of claim 21, wherein, to calculate the smoothed energy ratio, the one or more processors are configured to:

calculate the smoothed energy ratio according to the adjusted energy ratio as;

LF_EnergyRatio_sm⇐

(15·

LF_EnergyRatio_sm+Ratio)/16,

where LF_EnergyRatio_sm on the left side of the equation is the smoothed energy ratio of the current frame;

LF_EnergyRatio_sm on the right side of the equation is the smoothed energy ratio of a previous frame; and

Ratio is the adjusted energy ratio.

23. The encoder of claim 19, wherein the average normalized pitch correlation is obtained by:

calculating the average normalized pitch correlation as

Voicing=[R₁(P₁)+R₂(P₂)+R₃(P₃)+R₄(P₄)]/4,

where Voicing is the average normalized pitch correlation, R₁(P₁), R₂(P₂), R₃(P₃), and R₄(P₄) are four normalized pitch correlations calculated for four respective subframes of the current frame, and P₁, P₂, P₃, and P₄ are four pitch candidates, found in a pitch range from PIT_MIN to PIT_MAX, for the four respective subframes, wherein PIT_MIN is the first minimum pitch limitation, and PIT_MAX is a pitch limitation greater than the first minimum pitch limitation.

24. The encoder of claim 23, wherein each normalized pitch correlation is calculated according to:

25. The encoder of claim 23, wherein, to determine the value of the initial pitch lag candidate, the one or more processors are configured to:

determine the value of the initial pitch lag candidate as;

R(Pitch_Tp)=MAX{R(P),P=PIT_MIN0, . . . ,PIT_MIN}

where R(P) is a normalized pitch correlation for a pitch lag P, Pitch_Tp is the value of the initial pitch lag candidate;

PIT_MIN0 is the second minimum pitch limitation; and

PIT_MIN is the first minimum pitch limitation.

26. The encoder of claim 25, wherein the normalized pitch correlation is calculated according to:

27. The encoder of claim 18, wherein to determine the initial pitch lag candidate is a final pitch lag when the lack of low frequency energy in the current frame is detected, the one or more processors are configured to:

determine the initial pitch lag candidate is the final pitch lag when the lack of low frequency energy in the current frame is detected and a smoothed pitch correlation of the initial pitch lag candidate of the current frame is greater than a third threshold.

28. The encoder of claim 27, wherein the smooth pitch correlation is calculated according to:

Voicing0_sm⇐

(3·

Voicing0_sm+Voicing0)/4

where Voicing0_sm on the left side is the smoothed pitch correlation of the initial pitch lag candidate of the current frame;

Voicing0_sm on the right side is a smoothed pitch correlation of the initial pitch lag candidate of a previous frame; and

Voicing0 is equal to a normalized pitch correlation of the initial pitch lag candidate.

29. The encoder of claim 23, wherein, the first threshold is 35, and the second threshold is 50.

30. The encoder of claim 27, wherein to determine the initial pitch lag candidate is a final pitch lag when the lack of low frequency energy in the current frame is detected, the one or more processors are configured to:

determine the initial pitch lag candidate is the final pitch lag when the lack of low frequency energy in the current frame is detected, the smoothed pitch correlation of the initial pitch lag candidate of the current frame is greater than the third threshold, and the smoothed pitch correlation of the initial pitch lag candidate of the current frame is greater than a value of a fourth threshold multiplied by a smoothed pitch correlation Voicing_sm of the current frame.

31. The encoder of claim 30, wherein the smoothed pitch correlation Voicing_sm is calculated according to:

Voicing_sm⇐

(3·

Voicing_sm+Voicing)/4

where Voicing_sm on the left side is the smoothed pitch correlation of the current frame;

Voicing_sm on the right side is a smoothed pitch correlation of a previous frame; and

Voicing is the average normalized pitch correlation.

32. The encoder of claim 30, wherein the fourth threshold is 0.7.

33. The encoder of claim 18, wherein, the first minimum pitch limitation is 34 and the second minimum pitch limitation is 17 for 12.8 kilohertz (kHz) sampling frequency.

34. The encoder of claim 18, wherein, the one or more processors are further configured to:

encode the final pitch lag.