Method and apparatus for manipulating pitch and/or duration of a signal

• US 5,479,564 A
• Filed: 10/20/1994
• Issued: 12/26/1995
• Est. Priority Date: 08/09/1991
• Status: Expired due to Term

• Current Assignee: Nuance Communications, Inc. (Microsoft Corporation)
• Original Assignee: US Philips Corporation (Koninklijke Philips N.V.)
• Inventors: Eggen, Josephus H., Ma, Chang X., Vogten, Leonardus L. M., Verhelst, Werner D. E.
• Primary Examiner(s): MacDonald, Allen R.
• Assistant Examiner(s): Sartori, Michael A.
• Application Number:
  US08/326,791
• Time in Patent Office: 432 Days
• Field of Search: 395/2.67, 395/2.7, 395/2.76, 395/2.77, 395/2.74, 395/2.75, 381/36, 381/38, 381/40, 381/51, 381/53
• US Class Current: 704/267
• CPC Class Codes: G10L 21/04 Time compression or expansion

Abstract

Method and apparatus for manipulating an input signal (e.g. an audio (equivalent) signal) to obtain an output signal having a different pitch and/or duration. The method includes (a) positioning a chain of successive overlapping time windows with respect to the input signal; (b) deriving segments signals from the input signal and the windows; and (c) synthesizing the output signal by chained superposition of the segments signals. Each of the windows (except for the first window in the chain) is positioned by incrementing a position of the window from a corresponding position of a preceding window in the chain by a time interval. The time interval is substantially equal to a local pitch period for a portion of the input signal with respect to which the window is positioned. Accordingly, each of the windows of the chain (except for the first window) is positioned so that it begins at a predetermined time interval from a preceding window in the chain. The apparatus includes units for carrying out each of these processes.

Claims

1. A method of manipulating an input signal to obtain an output signal having a different pitch and/or duration than the input signal, the method comprising:

positioning a chain of successive overlapping time windows with respect to the input signal, each of the windows, except for a first window in the chain, being positioned by incrementing a position of that window from a corresponding position of a preceding window in the chain by a time interval which is substantially equal to a local pitch period for a portion of the input signal with respect to which that window will be positioned, said incrementing thereby determining where that window is positioned;

deriving segment signals from the input signal and the windows, each of the segment signals being derived by weighting the input signal as a function of position in a corresponding one of the windows; and

synthesizing the output signal by chained superposition of the segment signals.

2. The method according to claim 1, wherein the input signal is an audio signal and the method further comprises determining the local pitch period from the audio signal.

3. The method according to claim 2, wherein the local pitch period is determined by maximizing a measure of correlation between the audio signal and the audio signal shifted in time.

4. The method according to claim 2, wherein the local pitch period is determined using a position of a peak amplitude in a frequency spectrum of the audio signal.

5. The method according to claim 2, wherein the audio signal includes speech information with a stretch of unvoiced speech interposed between adjacent stretches of voiced speech, and the local pitch period for the stretch of unvoiced speech is determined by interpolating from local pitch periods determined for the adjacent stretches of voiced speech.

6. The method according to claim 1, further comprising manipulating the input signal so that the input signal has substantially uniform local pitch periods.

7. The method according to claim 1, further comprising deriving the input signal on the basis of overlapping an end portion of a first signal and a beginning portion of a second signal so that the beginning portion of the second signal begins at a position in time relative to the end portion of the first signal which minimizes a criteria which is indicative of a transition phenomenon in the output signal.

8. The method according to claim 7, wherein in deriving the input signal interpolation is performed with respect to the end portion of the first signal and the beginning portion of the second signal.

9. The method as claimed in claim 7, wherein the first signal and the second signal are audio signals and local pitch periods are determined from the first signal and the second signal.

10. The method as claimed in claim 7, further comprising manipulating the first signal and the second signal so that they both have substantially uniform local pitch periods.

11. The method as claimed in claim 1, wherein the output signal is synthesized by using each of the segment signals once.

12. The method as claimed in claim 1, wherein the windows have lengths which are independent of the change in pitch and/or duration between the output signal and the output signal.

13. An apparatus for manipulating an input signal to obtain an output signal having a different pitch and/or duration than the input signal, the apparatus comprising:

positioning means for positioning a chain of successive overlapping time windows with respect to the input signal;

incrementing means for determining a position of each of the windows, except for a first window in the chain, by incrementing from a corresponding position of a preceding window in the chain by a time interval which is substantially equal to a local pitch period for a portion of the input signal with respect to which that window will be positioned;

segmenting means for deriving segment signals from the input signal and the windows, each of the segment signals being derived by weighting the input signal as a function of position in a corresponding one of the windows; and

combination means for synthesizing the output signal by chained superposition of the segment signals.

14. The apparatus as claimed in claims 13, further comprising determining means for determining the local pitch period.

15. The apparatus as claimed in claims 13, further comprising derivation means for deriving the input signal on the basis of overlapping an end portion of a first signal and a beginning portion of a second signal, said derivation means being adapted to begin the beginning portion of the second signal at a position in time relative to the end portion of the first signal which minimizes a criterion which is indicative of a transition phenomenon in the output signal.

16. The apparatus according to claim 15, further comprising interpolation means for performing an interpolation with respect to the end portion of the first signal and the beginning portion of the second signal.

17. The apparatus as claimed in claim 13, wherein said combination means synthesizes the output signal by using each of the segment signals once.

18. The apparatus as claimed in claim 13, wherein the windows have lengths which are independent of the change in pitch and/or duration between the output signal and the input signal.

19. A method for producing an output signal from a first signal and a second signal, the method comprising:

overlapping the first and second signals so that a beginning portion of the second signal overlaps an end portion of the first signal, the beginning portion of the second signal beginning at a position in time relative to the end portion of the first signal which minimizes a criteria which is indicative of a transition phenomenon in the output signal;

positioning a chain of successive overlapping time windows with respect to the first and second signals, each of the windows, except for a first window in the chain, being positioned by incrementing a position of the that window from a corresponding position of a preceding window in the chain by a time interval which is substantially equal to a local pitch period for a portion of the first signal, the second signal or a combination of the first and second signals with respect to which that window will be positioned, said incrementing thereby determining where that window is positioned;

deriving segment signals from the first and second signals and the windows, each of the segment signals being derived by weighting the first signal, the second signal or a combination of the first and second Signals as a function of position in a corresponding one of the windows; and

synthesizing the output signal by chained superposition of the segment signals.

20. The method according to claim 19, further comprising performing an interpolation with respect to the end portion of the first signal and the beginning portion of the second signal.

21. The method as claimed in claim 19, wherein the first signal and the second signal are audio signals, and the method further comprises determining the local pitch periods from the first signal, the second signal or a combination of the first and second signals.

22. The method as claimed in claim 19, further comprising manipulating the first signal and the second signal so that the first signal and the second signal both have substantially uniform local pitch periods.

23. The method as claimed in claim 19, wherein the output signal is synthesized by using each of the segment signals once.

24. The method as claimed in claim 19, wherein the windows have lengths which are independent of the change in pitch and/or duration between the output signal and the input signal.

25. An apparatus for producing an output signal from a first signal and a second signal, the apparatus comprising:

overlapping means for overlapping the first and second signals so that a beginning portion of the second signal overlaps an end portion of the first signal, said overlapping means being adapted to position the beginning portion of the second signal at a position in time relative to the end portion of the first signal which minimizes a criteria which is indicative of a transition phenomenon in the output signal;

positioning means for positioning a chain of successive overlapping time windows with respect to the first and second signals;

incrementing means for determining a position of each of the windows, except for the first window in the chain, by incrementing from a corresponding position of a preceding window in the chain by a time interval which is substantially equal to a local pitch period for a portion of the first signal, the second signal or a combination of the first and second signals with respect to which that window will be positioned;

segmenting means for deriving segment signals from the first and second signals and the windows, each of the segment signals being derived by weighting the first signal, the second signal or a combination of the first and second signals as a function of position in a corresponding one of the windows; and

combination means for synthesizing the output signal by chained superposition of the segment signals.

26. The apparatus as claimed in claim 25, further comprising interpolation means for performing an interpolation with respect to the end portion of the first signal and the beginning portion of the second signal.

27. The apparatus as claimed in claim 25, wherein said combination means synthesizes the output signal by using each of the segment signals once.

28. The apparatus as claimed in claim 25, wherein the windows have lengths which are independent of the change in pitch and/or duration between the output signal and the input signal.