SPEECH ADAPTATION IN SPEECH SYNTHESIS

• US 20110282668A1
• Filed: 05/14/2010
• Published: 11/17/2011
• Est. Priority Date: 05/14/2010
• Status: Active Grant

• Current Assignee: General Motors LLC (General Motors Company)
• Original Assignee: General Motors LLC (General Motors Company)
• Inventors: Chengalvarayan, Rathinavelu, Talwar, Gaurav, Stefan, Jeffrey M.
• Granted Patent: US 9,564,120 B2
• Time in Patent Office: Days
• US Class Current: 704/260
• CPC Class Codes: G10L 13/033 Voice editing, e.g. manipulating the voice of the synthesiser
• CPC Class Codes: G10L 2021/0135 Voice conversion or morphing

Abstract

A method of and system for speech synthesis. First and second text inputs are received in a text-to-speech system, and processed into respective first and second speech outputs corresponding to stored speech respectively from first and second speakers using a processor of the system. The second speech output of the second speaker is adapted to sound like the first speech output of the first speaker.

Claims

1. A method of speech synthesis, comprising the steps of:

(a) receiving first and second text inputs in a text-to-speech system;

(b) processing the first and second text inputs into respective first and second speech outputs corresponding to stored speech respectively from first and second speakers using a processor of the system; and

(c) adapting the second speech output of the second speaker to sound like the first speech output of the first speaker.

2. The method of claim 1, further comprising the steps of:

(d) outputting the first speech output of the first speaker; and

(e) outputting the adapted second speech output of the second speaker.

3. The method of claim 2, wherein the first speech output is a navigational instruction and the second speech output is a navigational variable.

4. The method of claim 3, wherein the navigational instruction is a directional maneuver and the navigational variable is a street name.

5. The method of claim 2, further comprising the step of (f) modifying models used in conjunction with processing the stored speech from the second speaker.

6. The method of claim 5, wherein step (f) includes modifying Hidden Markov Models.

7. The method of claim 1, wherein step (c) includes:

(c1) analyzing acoustic features of the first speech output for at least one speaker specific characteristic of the first speaker;

(c2) adjusting an acoustic feature filter used to filter acoustic features from the second speech output, based on the at least one speaker specific characteristic of the first speaker; and

(c3) filtering acoustic features from the second speech output using the filter adjusted in step (c2).

8. The method of claim 7, wherein step (c3) includes adjusting at least one parameter of a mel-frequency cepstrum filter including at least one of filter bank central frequencies, filter bank cutoff frequencies, filter bank bandwidths, filter bank shape, or filter gain.

9. The method of claim 7, wherein the at least one speaker specific characteristic includes at least one of vocal tract or nasal cavity related characteristics.

10. The method of claim 9, wherein the characteristics include at least one of length, shape, transfer function, formants, or pitch frequency.

11. A computer program product including instructions on a computer readable medium and executable by a computer processor of a speech synthesis system to cause the system to implement steps comprising:

(a) receiving first and second text inputs in a text-to-speech synthesis system;

(b) processing the first and second text inputs into respective first and second speech outputs corresponding to stored speech respectively from first and second speakers using a processor of the system; and

(c) adapting the second speech output of the second speaker to sound like the first speech output of the first speaker.

12. The product of claim 11, wherein step (c) includes:

(c1) analyzing acoustic features of the first speech output for at least one speaker specific characteristic of the first speaker;

(c2) adjusting an acoustic feature filter used to filter acoustic features from the second speech output, based on the at least one speaker specific characteristic of the first speaker; and

(c3) filtering acoustic features from the second speech output using the filter adjusted in step (c2).

13. A speech synthesis system, comprising:

a first source of text;

a second source of text;

a first speech database including pre-recorded speech from a first speaker;

a second speech database including pre-recorded speech from a second speaker;

a pre-processor to convert text into synthesizable output;

a processor to convert first and second text inputs from the first and second sources of text into respective first and second speech outputs corresponding to the pre-recorded speech respectively from the first and second speakers; and

a post-processor to adapt the second speech output of the second speaker to sound like the first speech output of the first speaker.

14. The system of claim 13, further comprising:

an acoustic interface to convert speech output into audio signals; and

a speaker to convert the audio signals to audible speech.

15. The system of claim 14, wherein the speaker outputs the first speech output of the first speaker, and outputs the adapted second speech output of the second speaker.

16. The system of claim 13, wherein the post-processor modifies models used in conjunction with processing the stored speech from the second speaker.

17. The system of claim 13, wherein the post-processor analyzes acoustic features of the first speech output for at least one speaker specific characteristic of the first speaker, adjusts an acoustic feature filter used to filter acoustic features from the second speech output, based on the at least one speaker specific characteristic of the first speaker, and filters acoustic features from the second speech output using the adjust filter.

18. The system of claim 17, wherein the post-processor adjusts at least one parameter of a mel-frequency cepstrum filter including at least one of filter bank central frequencies, filter bank cutoff frequencies, filter bank bandwidths, filter bank shape, or filter gain.