System and Method for producing voice files for an automated concatenated voice system

• US 5,758,323 A
• Filed: 01/09/1996
• Issued: 05/26/1998
• Est. Priority Date: 01/09/1996
• Status: Expired due to Term

• Current Assignee: Qwest Communications International Incorporated (Lumen Technologies, Inc.)
• Original Assignee: U S West Marketing Resources Group, Inc.
• Inventors: Case, Eliot M.
• Primary Examiner(s): Tung, Kee M.
• Application Number:
  US08/587,125
• Time in Patent Office: 868 Days
• Field of Search: 395/2.87, 395/2.79, 395/2.22, 395/2.67, 395/2.76, 395/226, 395/227, 379/67, 379/68, 379/71, 379/88, 704/278, 704/270, 704/213, 704/258, 704/267, 705/26, 705/27
• US Class Current: 704/278
• CPC Class Codes: G10L 13/027 Concept to speech synthesisers; Generation of natural phrases from machine-based concepts generation of parameters for speech synthesis out of text G10L13/08
• CPC Class Codes: G10L 13/07 Concatenation rules

Abstract

A method for producing a voice file for use in an automated concatenated voice system. The words and phrases to be used in the system are scripted in a staged script, and read by a voice talent. The recording of the staged script as read by the voice talent is processed and edited to produce a plurality of naturally sounding words and phrases which may be concatenated into voice messages. The edited words and phrases are stored in a composite voice file for use by an automated concatenated voice system.

Claims

1. A method for producing a natural sounding voice file for an automated concatenation voice system comprising:

identifying new words to be entered into the voice file;

scripting a staged script in which the new words are formulated into sentences;

recording the staged script as read by a voice talent to generate digital voice data;

adjusting the amplitude of the digital voice data such that the amplitude of the words are substantially the same;

editing the adjusted digital voice data to identify each of the new words; and

storing the new words into the voice file for use in the automated concatenation system.

2. The method of claim 1 wherein said voice file is a composite voice file for storing a plurality of words and phrases.

3. The method of claim 1 further including the step of practicing the reading of said staged script by the voice talent to assure that the reading of the staged script is natural and proper voice inflections are used.

4. The method of claim 1 wherein said step of scripting a staged script further includes the staging of the script using a computer program.

5. The method of claim 1 wherein said step of editing includes the step of editing in accordance with a predetermined set of rules.

6. The method of claim 1 further including the step of automatically playing back each new word in a voice message.

7. The method of claim 1 further including the step of offline testing of the new words together with words previously stored in the voice file in a similar situation as they will be used in said automated concatenation system.

8. The method of claim 1 wherein said automated concatenation system is an automated voice concatenation system for voice advertisements.

9. The method of claim 1 wherein the step of adjusting further comprises the steps of:

generating an average amplitude map of said digital voice data; and

adjusting the amplitude of the digital voice data as a function of said average amplitude map.

10. A method for producing natural sounding voice files for an automated concatenation voice system comprising:

identifying new words or phrases to be entered into the voice file;

scripting a staged script in which the new words and phrases are formulated into real sentences;

recording the staged script as read by a voice talent to generate a composite recording;

processing the composite recording to increase clarity and to match words and phrases that are currently stored in the voice file;

precision editing of the composite recording to isolate and to assign an identification number to each of the new words and phrases; and

storing the new words and phrases into the voice file for use in the automated concatenation system;

wherein said step of processing comprises the step of compressing words and phrases in the composite recording such that the amplitude of the words and phrases are substantially the same.

11. The method of claim 10 wherein said step of compressing comprises the step of peak amplitude clamping.

12. A method for producing natural sounding voice files for an automated concatenation voice system comprising:

identifying new words or phrases to be entered into the voice file;

scripting a staged script in which the new words and phrases are formulated into real sentences;

recording the staged script as read by a voice talent to generate a composite recording;

processing the composite recording to increase clarity and to match words and phrases that are currently stored in the voice file;

precision editing of the composite recording to isolate and to assign an identification number to each of the new words and phrases; and

storing the new words and phrases into the voice file for use in the automated concatenation system;

wherein said step of editing includes the step of editing in accordance with a predetermined set of rules; and

wherein said predetermined set of rules comprises;

a) reducing by 12 dB a breath sound of an isolated phrase when the isolated phrase is long enough for the voice talent to take a breath in the middle of the recording;

b) editing is to be made in the least conspicuous place;

c) editing is to be made as close as possible to a zero crossing of the sounding;

d) editing is to be made outside the word or phrase being edited;

e) editing from the end of one word or phrase to the beginning of the next word or phrase should attempt to keep a normal continuation of the velocity of the sound;

f) editing should be made approximately 0.02±

0.005 seconds before the start of an isolated word or phrase; and

g) editing should be made approximately 0.02±

0.005 seconds after the end of a word or phrase.

13. The method of claim 12 wherein said step of editing to keep a normal continuation of the velocity of the sound further comprises:

editing the beginnings of a word or phrase at a zero crossing and going in the zero to positive direction;

editing the ends of a word or phrase at a zero crossing and going in the negative to zero direction.

14. The method of claim 12 wherein said step of editing 0.02±

0.005 seconds before the word or phrase for a fricative sound is made approximately at the beginning of the fricative sound, and wherein said step of editing 0.02±

0.005 seconds after a word or phrase for a fricative sound is made approximately at the ending of the fricative sound.

15. A system for producing natural sounding concatented voice files for an automated concatenation system comprising:

means for converting a voiced sound to digital voice data;

a digital data storage for storing the digital voice data;

a generator for generating an average amplitude map of said digital voice data stored in the digital data storage;

a peak amplitude clamping processor to adjust the amplitude of the digital voice data to a predetermined target level using said average amplitude map such that each word and syllable has approximately the same amplitude;

a word and phrase editor for identifying words or phrases in said digital voice data and assigning them individual identification numbers;

a voice file for storing the words and phrases identified by the word and phrase editor.

16. The system of claim 15 further including an off-line test system for testing the edited words and phrases together with words and phrases stored in the voice file prior to storing the edited words and phrases in the voice file.

17. The system of claim 15 wherein said voice file is a composite voice file storing a plurality of words and phrases.