Computationally efficient sine wave synthesis for acoustic waveform processing

• US 4,937,873 A
• Filed: 04/08/1988
• Issued: 06/26/1990
• Est. Priority Date: 03/18/1985
• Status: Expired due to Term

• Current Assignee: Massachusetts Institute of Technology
• Original Assignee: Massachusetts Institute of Technology
• Inventors: Quatieri, Thomas F. Jr., McAulay, Robert J.
• Primary Examiner(s): Harkcom, Gary V.
• Assistant Examiner(s): Knepper, David D.
• Application Number:
  US07/179,528
• Time in Patent Office: 809 Days
• Field of Search: 381/29-40, 381/46, 381/47, 381/50-53, 381/94
• US Class Current: 704/265
• CPC Class Codes: G10L 19/093 using sinusoidal excitation models

Abstract

Methods and apparatus for reducing discontinuities between frames of sinusoidally modeled acoustic waveforms, such as speech, which occur when sampling at low frame rates. A Fast Fourier Transform-based overlap-add technique is applied to amplitude, frequency and phase components of sinusoidal waves after frame-to-frame sine wave matching has been performed. Matched sine wave amplitudes and frequencies are linearly interpolated and mid-point phase is estimated such that the mid-frame sine wave is best fit to the most recent half-frame segments of the lagging and leading sine waves. Synthetic mid-frame sine waves are generated using the interpolated amplitude and frequency and estimated phase values. Synthesized acoustic waveforms of high quality from original source waveforms can be produced in sinusoidal analysis/synthesis operations at coding frame rates of 50 Hz and lower. The methods and devices disclosed herein are particularly useful for computationally efficient coding and synthesis of speech waveforms.

Claims

1. A method of processing an acoustic waveform, the method comprising:

sampling a waveform to obtain a series of discrete samples and constructing therefrom a series of frames, each frame spanning a plurality of samples;

analyzing each frame of samples to extract a set of variable frequency components having individual amplitudes;

tracking said components from one frame to a next frame, said tracking including matching a component from the one frame with a component in the next frame having a similar value regardless of shifts in frequency and spectral energy; and

interpolating the values of the components from the one frame to the next frame by performing an overlap-and-add function utilizing Fourier analysis to generate a reconstruction of said waveforms.

2. The method of claim 1 wherein said interpolating step further includes estimating mid-frame values and interpolating between said mid-frame values and values obtained during each frame in order to generate a refined representation of the waveform.

3. The method of claim 2 wherein said estimating step further includes deriving mid-frame amplitude and frequency values by linear interpolation of lagging and leading sine waves.

4. The method of claim 2 wherein said estimating step further includes providing a mid-frame phase value such that the sine wave corresponding to the interpolated mid-frame values of the parametric representation is best fit to predetermined segments of lagging and leading sine waves.

5. The method of claim 2 wherein said estimating step further includes deriving mid-frame phase values from the lagging and leading sine waves according to the following equation

6. The method of claim 1 wherein the method further includes suppressing tonal noise values.

7. The method of claim 6 wherein the method further includes estimating a noise envelope and using said noise envelope estimate to drive a noise suppression filter.

8. The method of claim 6 wherein the method further includes generating broadband noise to replace said suppressed noise values.

9. A method for suppressing tonal noise artifacts during the reconstruction of an acoustic waveform from a sinusoidal parametric representation of the waveform, the method comprising;

estimating a noise envelope from a set of variable frequency components having individual amplitudes which comprise a parametric representation of the waveform;

reconstructing an acoustic waveform from said parametric representation; and

filtering said reconstructed waveform using said noise envelope estimates to suppress tonal noise estimates.

10. A method of deriving phase values for frequency components during reconstruction of an acoustic waveform from a sinusoidal representation of the waveform, the method comprising:

determining a phase of the fundamental frequency by integration of a pitch frequency obtained by linear interpolation of matched fundamental frequencies between successive frames;

determining a pitch onset time by locating the time at which the phase function crosses the nearest multiple of the phase synchrony point; and

allocating phase values to the frequency components, such that all of the frequency components come into phase every pitch onset time.

11. A system for processing an acoustic waveform, the system comprising

sampling means for sampling a waveform to obtain a series of discrete samples and constructing therefrom a series of frames, each frame spanning a plurality of samples,

analyzing means for analyzing each frame of samples to extract a set of variable frequency components having individual amplitudes,

tracking means for tracking said components from one frame to a next frame, said tracking means including matching means for matching a component from the one frame with a component in the next frame having a similar value regardless of shifts in frequency and special energy,

interpolating means for interpolating the values of the components from the one frame to the next frame, including means for performing an overlap-and-add function utilizing Fourier analysis to generate a reconstruction of said waveform.

12. The system of claim 11 wherein said interpolating means further includes mid-frame estimating means for estimating mid-frame values and means for interpolating between said mid-frames values and values obtained during each frame in order to generate a refined representation of the waveform.

13. The system of claim 12 wherein said mid-frame estimating means further includes means for linearly interpolating the amplitude and frequency values of the lagging and leading sin waves to obtain mid-frame values.

14. The system of claim 12 wherein said mid-frame estimating means further includes means for deriving mid-frame phase values such that sine waves corresponding to the interpolated mid-frame values of the parametric representation is best fit to predetermined segments of lagging and leading sine waves.

15. The system of claim 12 wherein said mid-frame estimating means further includes means for deriving mid-frame phase values from lagging and leading sine waves according to the following equation:

space="preserve" listing-type="equation">θ

(M)=(θ

.sub.o +θ

.sub.1)/2+(ω

.sub.o -ω

.sub.1)/2.N/4+π

M

where M is an integer whose value is chosen, such that π

M is closest to

space="preserve" listing-type="equation">(θ

.sub.o -θ

.sub.1)/2+(ω

.sub.o +ω

.sub.1)/2.N/4

and where θ

_o is the phase of the lagging frame, θ

₁ is the phase of the leading frame, ω

hd o is the frequency of the lagging frame, ω

₁ is the frequency of the leading frame, and N is the analysis frame length.

16. The system of claim 11 wherein said system further includes means for suppressing tonal values.

17. The system of claim 16 wherein said system further includes noise estimating means for estimating a noise envelope and a filter means for suppressing tonal noise values in response to said noise envelope estimate.

18. The system of claim 16 wherein said system further includes a broadband noise generator to replace said suppressed noise values with broadband noise.

19. A receiver for receiving a coded parametric representation of an acoustic waveform in which the representation comprises as set of variable frequency components having individual amplitudes defining sine waves which can be summed to recreate the waveform at a particular frame of time, the receiver comprising:

decoding means for extracting a set of frequency components having individual amplitudes from each frame of a coded representation of an acoustic waveform;

tracking means for tracking said components from one frame to a next frame, said tracking means, including matching means for matching a component from the one frame with a component in the next frame having a similar value regardless of shifts in frequency and spectral energy; and

interpolation means for interpolating the values of the components from the one frame to the next frame, including means for performing an overlap-and-add function utilizing Fourier analysis, to generate a reconstruction of said waveform.

20. The receiver of claim 19 wherein said interpolating means further includes mid-frame estimating means for estimating mid-frame values and means for interpolating between said mid-frames values and values obtained during each frame in order to generate a refined representation of the waveform.

21. The receiver of claim 20 wherein said mid-frame estimating means further includes means for linearly interpolating the amplitude and frequency values of the lagging and leading sine waves to obtain mid-frame values.

22. The receiver of claim 20 wherein said mid-frame estimating means further includes means for deriving mid-frame phase values such that sine waves corresponding to the interpolated mid-frame values of the parametric representation is best fit to predetermined segments of lagging and leading sine waves.

23. The receiver of claim 20 wherein said mid-frame estimating means further includes means for deriving mid-frame phase values from lagging and leading sine waves according to the following equation:

space="preserve" listing-type="equation">θ

(M)=(θ

.sub.o +θ

.sub.1)/2+(ω

.sub.o -ω

.sub.1)/2.N/4+π

M

where M is an integer whose value is chosen, such that π

M is closest to

space="preserve" listing-type="equation">(θ

.sub.o -θ

.sub.1)/2+(ω

.sub.o +ω

.sub.1)/2.N/4

and where θ

_o is the phase of the lagging frame, θ

₁ is the phase of the leading frame, ω

_o is the frequency of the lagging frame, ω

₁ is the frequency of the leading frame, and N is the analysis frame length.

24. The receiver of claim 19 wherein said system further includes means for suppressing tonal values.

25. The receiver of claim 24 wherein said system further includes noise estimating means for estimating a noise envelope and a filter means for suppressing tonal noise values in response to said noise envelope estimate.

26. The receiver of claim 24 wherein said system further includes a broadband noise generator to replace said suppressed noise values with broadband noise.