Processing device for speech synthesis by addition of overlapping wave forms

• US 5,524,172 A
• Filed: 04/04/1994
• Issued: 06/04/1996
• Est. Priority Date: 09/02/1988
• Status: Expired due to Term

• Current Assignee: Represented By The Ministry of Posts Telecommunications and Space Centre, Represented By The Ministry of Posts Telecommunications and Space Centre National D'Etudes Des Telecommunicationss
• Original Assignee: Represented By The Ministry of Posts Telecommunications and Space Centre National D'Etudes Des Telecommunicationss
• Inventors: Hamon, Christian
• Primary Examiner(s): MacDonald, Allen R.
• Assistant Examiner(s): VILLAMAR, CARLOS
• Application Number:
  US08/224,652
• Time in Patent Office: 792 Days
• Field of Search: 381/50-52, 395/2.67-2.78
• US Class Current: 704/268
• CPC Class Codes: G10L 13/07 Concatenation rules

Abstract

A process of speech synthesis by the domain overlap-addition of elements stored in a dictionary as waveforms, comprises supplying a sequence of phoneme codes and respective prosodic information, and, for each phoneme, analyzing and synthesizing each phoneme, and then concatenating the synthesized phonemes. For each phoneme, two diphones are selected among the stored diphones and the presence of voicing is determined. For voiced phonemes, the respective waveforms of the two diphones constituting the phoneme are filtered by a window which is centered on a point of the selected waveform representative of the beginning of a pulse response of vocal cords to excitation thereof. The window has a width substantially equal to twice the greater of the original fundamental period or the fundamental synthesis period and has an amplitude progressively decreasing from the center of the window. The signals resulting from the filtering and obtained for each diphone are time shifted so as to be spaced apart by a time equal to the fundamental synthesis period.

Claims

1. Method of speech synthesis from speech sound elements comprising the steps of:

(a) analyzing at least voiced sounds of the sound element, by windowing by means of a filtering window having an amplitude decreasing to zero at the edges of the window, whose width is at least substantially equal to the shorter of an original fundamental period and a fundamental synthesis period,

(b) replacing the signals resulting from windowing corresponding to each sound element with a time shift thereof equal to the fundamental synthesis period, which is lesser than or greater than the original fundamental period responsive to prosodic information relative to the fundamental synthesis period, and

(c) summing the thus shifted signal to synthesize speech, said method being devoid of a modification of a pitch period of the speech sounds elements by spectral transformation between steps (a) and (b).

2. Method according to claim 1, comprising the step of decreasing speech frequency by selecting the width of the window as substantially equal to twice the original fundamental period.

3. A method according to claim 1, comprising the step of reducing speech frequency, wherein the width of the window is substantially equal to twice the original voicing period.

4. Method of speech synthesis from sound elements stored in a dictionary of waveforms, for speech conversion, consisting of the following steps:

(a) analyzing an original speech signal, said analysis including, at least for voiced sounds, subjecting the respective waveforms of the respective sound elements to filtering by windows, each of said windows having a width at least substantially equal to twice the lesser of an original fundamental period or a fundamental synthesis period and having an amplitude progressively decreasing from the center of the window to zero at the edges thereof,

(b) replacing the signals resulting from said filtering with such a time shift that said signals are spaced apart by a time equal to the fundamental synthesis period, and

(c) adding the replaced signals for synthesis of speech.

5. Method according to claim 4 comprising the step of decreasing a speech frequency by selecting the width of the window as substantially equal to twice the original fundamental period.

6. A method according to claim 4, comprising the step of reducing speech frequency, wherein the width of the window is substantially equal to twice the original voicing period.

7. A method of speech synthesis by time domain overlap addition of waveforms comprising the steps of analyzing at least voiced sounds of an original signal by weighting said original signal with windows synchronous with the voicing or pitch periods of said original signal stored as waveforms, to produce windowed waveforms, and directly repositioning said windowed waveforms for synthesis by mutual addition with a time interval therebetween which is lesser or greater than an original interval depending on prosodic information, wherein said windows each have an amplitude progressively decreasing to zero at the edges of the window and a width which is at least substantially equal to twice the shorter of a original voicing period or twice a synthesis voicing period.

8. A method according to claim 7, comprising a preliminary step of computing and storing said waveforms in a dictionary of diphones.

9. A method according to claim 7, wherein each said window is approximately centered on the beginning of a pulse response of the vocal tract to an excitation of the vocal cords for the respective waveform.

10. A method according to claim 7, wherein the windows are Hanning windows.

11. A method according to claim 7 comprising the step of increasing speech frequency, wherein the width of the window is substantially equal to twice the synthesis period.

12. A method according to claim 7, comprising the step of reducing speech frequency, wherein the width of the window is substantially equal to twice the original voicing period.