Method and configuration for determining a descriptive feature of a speech signal

• US 20020035469A1
• Filed: 09/10/2001
• Published: 03/21/2002
• Est. Priority Date: 03/08/1999
• Status: Active Grant

• Current Assignee: Siemens AG
• Original Assignee: Siemens AG
• Inventors: Holzapfel, Martin
• Granted Patent: US 6,523,005 B2
• Time in Patent Office: Days
• US Class Current: 704/211
• CPC Class Codes: G10L 15/02 Feature extraction for speech recognition; Selection of recognition unit
• CPC Class Codes: G10L 15/063 Training
• CPC Class Codes: G10L 2015/0635 updating or merging of old and new templates; Mean values; Weighting

Abstract

A method and also a configuration for determining a descriptive feature of a speech signal, in which a first speech model is trained with a first time pattern and a second speech model is trained with a second time pattern. The second speech model is initialized with the first speech model.

Claims

1. A method for determining a descriptive feature of a speech signal, which comprises:

training a first speech model with a first time pattern;

training a second speech model with a second time pattern; and

initializing the second speech model with the first speech model.

2. The method according to claim 1, which comprises providing the second time pattern to he smaller than the first time pattern.

3. The method according to claim 1, wherein the step of initializing the second speech model includes initializing the second speech model with segment boundaries of the first speech model.

4. The method according to claim 1, which comprises providing the first speech model as a hidden Markov model.

5. The method according to claim 4, which comprises determining segment boundaries of the first speech model by training the hidden Markov model.

6. The method according to claim 1, which comprises providing the second speech model as a hidden Markov model.

7. The method according to claim 1, which comprises providing a descriptive feature as a feature vector.

8. The method according to claim 1, which comprises situating the first time pattern in a range between 4 ms and 10 ms.

9. The method according to claim 1, which comprises situating the second time pattern in a range greater than zero and running up to 6 ms.

10. The method according to claim 1, which comprises performing the training steps and the initialization step in a system for speech processing.

11. The method according to claim 1, which comprises performing the training steps and the initialization step in a system for speech synthesis.

12. A configuration for determining a descriptive feature of a speech signal, comprising a processor unit configured for:

training a first speech model with a first time pattern;

training a second speech model with a second time pattern; and

initializing the second speech model with the first speech model.