Method and apparatus to modify pitch estimation function in acoustic signal musical note pitch extraction

• US 20060065107A1
• Filed: 09/24/2004
• Published: 03/30/2006
• Est. Priority Date: 09/24/2004
• Status: Active Grant

• Current Assignee: Nokia Corporation
• Original Assignee: Nokia Corporation
• Inventors: Kosonen, Timo Antero
• Granted Patent: US 7,230,176 B2
• Time in Patent Office: Days
• US Class Current: 84/616
• CPC Class Codes: G10H 1/0008 Associated control or indicating means
• CPC Class Codes: G10H 2210/066 for pitch analysis as part of wider processing for musical purposes, e.g. transcription, musical performance evaluation; Pitch recognition, e.g. in polyphonic sounds; Estimation or use of missing fundamental
• CPC Class Codes: G10H 2210/086 for transcription of raw audio or music data to a displayed or printed staff representation or to displayable MIDI-like note-oriented data, e.g. in pianoroll format
• CPC Class Codes: G10H 2230/015 PDA [personal digital assistant] or palmtop computing devices used for musical purposes, e.g. portable music players, tablet computers, e-readers or smart phones in which mobile telephony functions need not be used touch-screen interfaces G10H2220/096
• CPC Class Codes: G10H 2230/021 Mobile ringtone, i.e. generation, transmission, conversion or downloading of ringing tones or other sounds for mobile telephony; Special musical data formats or protocols herefor mobile telephone transmission specifically adapted for electrophonic musical tools or instruments G10H2240/251
• CPC Class Codes: G10H 2250/161 Logarithmic functions, scaling or conversion, e.g. to reflect human auditory perception of loudness or frequency

Abstract

In one aspect thereof this invention provides a method to estimate pitch in an acoustic signal. The method includes initializing a function ƒ_t and a time t, where t=0, x′₀=ƒ₀(F₀), x′₀ is a pitch estimate at time zero and F₀ is a frequency of the acoustic signal at time zero; determining at least one pitch estimate using the function x′_t=ƒ_t(F_t) by an iterative process of creating ƒ_t+1(F_t+1) based at least partly on pitch estimates x′_t, x′_t−1, x′_{t−2, x′t−3}, . . . , and functions ƒ_t(F_t), ƒ_t−1(F_t−1), ƒ_t−2(F_t−2), ƒ_t−3(F_t−3) . . . and incrementing t; and calculating at least one final pitch estimate. Embodiments of this invention can be applied to pitch extraction with various different input acoustic signal characteristics, such as just intonation, pitch shift in the frequency domain, and non-12-step-equal-temperament tuning.

Claims

1. A method to estimate pitch in an acoustic signal, comprising:

initializing a function ƒ

_t and a time t, where t=0, x′

₀=ƒ

₀(F₀), x′

₀ is a pitch estimate at time zero and F₀ is a frequency of the acoustic signal at time zero;

determining at least one pitch estimate using the function x′

_t=ƒ

_t(F_t) by an iterative process of creating ƒ

_t+1(F_t+1) based at least partly on pitch estimates x′

_t, x′

<sub>t−

1</sub>, x′

<sub>t−

2</sub>,x<sub>t−

3</sub>. . . and functions ƒ

_t(F_t), ƒ

<sub>t−

1</sub>(F<sub>t−

1</sub>), ƒ

<sub>t−

2</sub>(F<sub>t−

2</sub>), ƒ

<sub>t−

3</sub>(F<sub>t−

3</sub>) . . . and incrementing t; and

calculating at least one final pitch estimate.

2. A method as in claim 1, where x′

_t=ƒ

(F_t) is represented by x′

_t=m+s*log₂(F_t/F_b), where s defines a number of notes in an octave, and F_b is a reference frequency.

3. A method as in claim 2, and for a case of just intonation, the method further comprising setting s=12, m=a MIDI number of a root note in the key, F_b=440×

2<sup>((m−

69)/12)</sup> Hz, and mapping the ratio F_t/F_b to an adjusted ratio R_t.

4. A method as in claim 3, where mapping comprises using a table comprising:

F_t/F_b R_t . . . . . . 2<sup>(−

1)</sup> ×

9/5 2<sup>(−

2)/12</sup> 2<sup>(−

1)</sup> ×

15/8 2<sup>(−

1)/12</sup> 2⁰ ×

1 2^0/12 2⁰ ×

16/15 2^1/12 2⁰ ×

9/8 2^2/12 2⁰ ×

6/5 2^3/12 2⁰ ×

5/4 2^4/12 2⁰ ×

4/3 2^5/12 2⁰ ×

45/32 2^6/12 2⁰ ×

3/2 2^7/12 2⁰ ×

8/5 2^8/12 2⁰ ×

5/3 2^9/12 2⁰ ×

9/5 2^10/12 2⁰ ×

15/8 2^11/12 2¹ ×

1 2^12/12 2¹ ×

16/15 2^13/12 2¹ ×

9/8 2^14/12 2¹ ×

6/5 2^15/12 2¹ ×

5/4 2^16/12 2¹ ×

4/3 2^17/12 . . . . . .

5. A method as in claim 2, further comprising, subsequent to calculating multiple final pitch estimates x_t,i of a first note:

setting m=x_t where x_t depends on all x_t,i and modifying F_b to be a corresponding frequency;

continuing the iterative process; and

mapping the ratio F_t/F_b to an adjusted ratio R_t for each note according to;

F_t/F_bR_t......2<sup>(−

1)</sup> ×

9/52<sup>(−

2)/12</sup>2<sup>(−

1)</sup> ×

15/82<sup>(−

1)/12</sup>2⁰ ×

12^0/122⁰ ×

16/152^1/122⁰ ×

9/82^2/122⁰ ×

6/52^3/122⁰ ×

5/42^4/122⁰ ×

4/32^5/122⁰ ×

45/322^6/122⁰ ×

3/22^7/122⁰ ×

8/52^8/122⁰ ×

5/32^9/122⁰ ×

9/52^10/122⁰ ×

15/82^11/122¹ ×

12^12/122¹ ×

16/152^13/122¹ ×

9/82^14/122¹ ×

6/52^15/122¹ ×

5/42^16/122¹ ×

4/32^17/12......

6. A method as in claim 5, where during the iterative process initial pitch estimates are computed as x′

_t=m+s*log₂(R_t).

7. A method as in claim 1, where x′

_t=m+s*log₂(R_t), where s=12 and R_t=(F_t+(delta))/F_b to accommodate a shift in pitch.

8. A method as in claim 1, where x′

_t=m+s*log₂(F_t/F_b), where s=(alpha) * 12, where the value of (alpha) defines by how much a musical scale is contracted or expanded, and where values of m and F_b are selected to be from a range of pitch frequencies that are known to be in tune.

9. A method as in claim 1, where x′

_t=s*log₂(R_t), where R_t depends on F_t and F_b, and where s defines a number of steps in one octave.

10. A method as in claim 9, where R_t=F_t/F_b for a case of equal tuning.

11. A method as in claim 9, where R_t represents a mapping of F_t/F_b for a case of non-equal tuning.

14. A computer-readable storage medium as in claim 3, and for a case of just intonation, the method further comprising setting s=12, m=a MIDI number of a root note in the key, F_b=440×

2<sup>((m−

69)/12)</sup> Hz, and mapping the ratio F_t/F_b to an adjusted ratio R_t.

15. A computer-readable storage medium as in claim 14, where mapping comprises using a table comprising:

F_t/F_b R_t . . . . . . 2<sup>(−

1)</sup> ×

9/5 2<sup>(−

2)/12</sup> 2<sup>(−

1)</sup> ×

15/8 2<sup>(−

1)/12</sup> 2⁰ ×

1 2^0/12 2⁰ ×

16/15 2^1/12 2⁰ ×

9/8 2^2/12 2⁰ ×

6/5 2^3/12 2⁰ ×

5/4 2^4/12 2⁰ ×

4/3 2^5/12 2⁰ ×

45/32 2^6/12 2⁰ ×

3/2 2^7/12 2⁰ ×

8/5 2^8/12 2⁰ ×

5/3 2^9/12 2⁰ ×

9/5 2^10/12 2⁰ ×

15/8 2^11/12 2¹ ×

1 2^12/12 2¹ ×

16/15 2^13/12 2¹ ×

9/8 2^14/12 2¹ ×

6/5 2^15/12 2¹ ×

5/4 2^16/12 2¹ ×

4/3 2^17/12 . . . . . .

12. A computer-readable storage medium storing a computer program for causing the computer to estimate pitch in an acoustic signal by operations of:

initializing a function ƒ

_t and a time t, where t=0, x′

₀=ƒ

₀(F₀), x′

₀ is a pitch estimate at time zero and F₀ is a frequency of the acoustic signal at time zero;

determining at least one pitch estimate using the function x′

_t=ƒ

_t(F_t) by an iterative process of creating ƒ

_t+1(F_t+1) based at least partly on pitch estimates x′

_t, x′

<sub>t−

1</sub>, x<sub>t−

2</sub>, x<sub>t−

3</sub>, . . . , and functions ƒ

_t(F_t), ƒ

<sub>t−

1</sub>(F<sub>t−

1</sub>), ƒ

<sub>t−

2</sub>(F<sub>t−

2</sub>), ƒ

<sub>t−

3</sub>(F<sub>t−

3</sub>) . . . and incrementing t; and

calculating at least one final pitch estimate.

13. A computer-readable storage medium as in claim 12, where x′

_t=ƒ

(F_t) is represented by x′

_t=m+s*log₂(F_t/F_b), where s defines a number of notes in an octave, and F_b is a reference frequency.

16. A computer-readable storage medium as in claim 13, further comprising, subsequent to calculating multiple final pitch estimates x_t,i of a first note:

setting m=x_t, where x_t depends on all x_t,i, and modifying F_b to be a corresponding frequency;

continuing the iterative process; and

mapping the ratio F_t/F_b to an adjusted ratio R_t for each note according to;

F_t/F_bR_t......2<sup>(−

1)</sup> ×

9/52<sup>(−

2)/12</sup>2<sup>(−

1)</sup> ×

15/82<sup>(−

1)/12</sup>2⁰ ×

12^0/122⁰ ×

16/152^1/122⁰ ×

9/82^2/122⁰ ×

6/52^3/122⁰ ×

5/42^4/122⁰ ×

4/32^5/122⁰ ×

45/322^6/122⁰ ×

3/22^7/122⁰ ×

8/52^8/122⁰ ×

5/32^9/122⁰ ×

9/52^10/122⁰ ×

15/82^11/122¹ ×

12^12/122¹ ×

16/152^13/122¹ ×

9/82^14/122¹ ×

6/52^15/122¹ ×

5/42^16/122¹ ×

4/32^17/12......

17. A computer-readable storage medium as in claim 16, where during the iterative process initial pitch estimates are computed as x′

_t=m+s*log₂(R_t).

18. A computer-readable storage medium as in claim 12, where x′

_t=m+s*log₂(R_t), where s=12 and R_t=(F_t+(delta))/F_b to accommodate a shift in pitch.

19. A computer-readable storage medium as in claim 12, where x′

_t=m+s*log₂(F_t/F_b), where s=(alpha)*12, where the value of (alpha) defines by how much a musical scale is contracted or expanded, and where values of m and F_b are selected to be from a range of pitch frequencies that are known to be in tune.

20. A computer-readable storage medium as in claim 12, where x′

_t=s*log₂(R_t), where R_t depends on F_t and F_b, and where s defines a number of steps in one octave.

21. A computer-readable storage medium as in claim 20, where R_t=F_t/F_b for a case of equal tuning.

22. A computer-readable storage medium as in claim 20, where R_t=to a mapping of F_t/F_b for a case of non-equal tuning.

23. A system comprising means for receiving data representing an acoustic signal and processing means to process the received data to estimate a pitch of the acoustic signal, where said processing means comprises means for initializing a function ƒ

_t, and a time t, where t=0, x′

₀=ƒ

₀(F₀), x′

₀ is a pitch estimate at time zero and F₀ is a frequency of the acoustic signal at time zero;

means for determining at least one pitch estimate using the function x′

_t=ƒ

_t(F₀) by an iterative process of creating ƒ

_t+1(F_t+1) based at least partly on pitch estimates x′

_t, x′

<sub>t−

1</sub>, x′

<sub>t−

2</sub>, x′

<sub>t−

3</sub>, . . . , and functions ƒ

_t(F_t), ƒ

<sub>t−

1</sub>(F<sub>t−

1</sub>), ƒ

<sub>t−

2</sub>(F<sub>t−

2</sub>), ƒ

<sub>t−

3</sub>(F<sub>t−

3</sub>) . . . and incrementing t; and

means for determining at least one final pitch estimate (x_t).

24. A system as in claim 23, where said receiving means comprises a receiver means having an input coupled to a data communications network.

25. A system as in claim 23, where said receiving means comprises an acoustic transducer means and an analog to digital conversion means for converting an acoustic signal to data that represents the acoustic signal.

26. A system as in claim 23, where the acoustic signal comprises a person'"'"'s voice.

27. A system as in claim 26, where the system comprises a telephone, where the processor means uses the at least one final pitch estimate for generating a ringing tone.

28. A system as in claim 23, where determining the final pitch estimate (x_t) determines a final pitch estimate of a single note from multiple pitch estimates (x_t,i) that have been determined for the same note.

29. A system as in claim 28, where at least for a case of a loss of a sense of key, the final pitch estimate, which depends on all x_t,i, is determined for a note before a recursion may continue for a next note with a slightly or clearly different key.

30. A system as in claim 28, where determining final pitch estimate comprises a shifting operation that adds a histogram mass center c_t to a result of the pitch estimation.