Methods for generating the voiced portion of speech signals

• US 5,195,166 A
• Filed: 11/21/1991
• Issued: 03/16/1993
• Est. Priority Date: 09/20/1990
• Status: Expired due to Term

• Current Assignee: Digital Voice Systems, Inc.
• Original Assignee: Digital Voice Systems, Inc.
• Inventors: Hardwick, John C., Lim, Jae S.
• Primary Examiner(s): Fleming, Michael R.
• Assistant Examiner(s): Doerrler, Michelle
• Application Number:
  US07/795,963
• Time in Patent Office: 481 Days
• Field of Search: 381/29-53, 395/2
• US Class Current: 704/200
• CPC Class Codes: G10L 19/087 using mixed excitation models, e.g. MELP, MBE, split band LPC or HVXC
• CPC Class Codes: G10L 25/90 Pitch determination of speech signals
• CPC Class Codes: G10L 25/93 Discriminating between voiced and unvoiced parts of speech signals G10L25/90 takes precedence

Abstract

The pitch estimation method is improved. Sub-integer resolution pitch values are estimated in making the initial pitch estimate; the sub-integer pitch values are preferably estimated by interpolating intermediate variables between integer values. Pitch regions are used to reduce the amount of computation required in making the initial pitch estimate. Pitch-dependent resolution is used in making the initial pitch estimate, with higher resolution being used for smaller values of pitch. The accuracy of the voiced/unvoiced decision is improved by making the decision dependent on the energy of the current segment relative to the energy of recent prior segments; if the relative energy is low, the current segment favors an unvoiced decision; if high, it favors a voiced decision. Voiced harmonics are generated using a hybrid approach; some voiced harmonics are generated in the time domain, whereas the remaining harmonics are generated in the frequency domain; this preserves much of the computational savings of the frequency domain approach, while at the same time improving speech quality. Voiced harmonics generated in the frequency domin are generated with higher frequency accuracy; the harmonics are frequency sealed, transformed into the time domain with a Discrete Fourier Transform, interpolated and then time scaled.

Claims

1. A method for generating the voiced portion of a speech signal of the type generated by synthesis from voiced harmonics, the method comprising the steps of:

receiving a signal containing information on a plurality of voiced harmonics, including information on first and second groups of said voiced harmonics;

generating said first group of voiced harmonics using a time domain synthesis method;

generating said second group of voiced harmonics using a frequency domain synthesis method; and

combining said generated first and second groups of voiced harmonics to produce said voiced portion of a speech signal.

2. The method of claim 1 wherein said first group comprises low-frequency harmonics.

3. The method of claim 1 or 2 wherein said second group comprises high-frequency harmonics.

4. The method of claim 3 wherein said time domain synthesis is performed by generating a low-order piecewise phase polynomial.

5. The method of claim 3 wherein said frequency domain synthesis is performed using the method comprising the steps of:

linearly frequency scaling said information on said voiced harmonics according to the mapping ω

₀ →

2π

/L, where L is some small integer, to generate frequency-scaled harmonics;

performing an L-point Inverse Discrete Fourier Transform (DFT) to simultaneously transform said frequency scaled harmonics into the time domain; and

performing interpolation and time scaling to generate said second group of voiced harmonics.

6. The method of claim 1 wherein said time domain synthesis is performed by generating a low-order piecewise phase polynomial.

8. The method of claim 5 or 7 wherein said DFT is computed with a Fast Fourier Transform, and L is a highly composite number.

9. The method of claim 5 or 7 wherein said interpolation is performed with linear interpolation.

7. A method for generating the voiced portion of a speech signal of the type generated by synthesis from voiced harmonics, the method comprising the steps of:

receiving a signal containing information on a plurality of voiced harmonics;

linearly frequency scaling said information on said voiced harmonics according to the mapping ω

₀ →

2π

/L, where L is some small integer, to generate frequency-scaled harmonics;

performing an L-point Inverse Discrete Fourier Transform (DFT) to simultaneously transform said frequency scaled harmonics into the time domain;

performing interpolation and time scaling to generate said plurality of voiced harmonics; and

combining said voiced harmonics to produce said voiced portion of a speech signal.