Process of voice recognition in a harsh environment, and device for implementation

• US 6,128,594 A
• Filed: 09/26/1997
• Issued: 10/03/2000
• Est. Priority Date: 01/26/1996
• Status: Expired due to Term

• Current Assignee: Sextant Avionique
• Original Assignee: Sextant Avionique
• Inventors: Pastor, Dominique, Reynaud, Gerard, Gulli, Christian
• Primary Examiner(s): Dorvil, Richemond
• Assistant Examiner(s): Lerner, Martin
• Application Number:
  US08/913,849
• Time in Patent Office: 1,103 Days
• Field of Search: 704/10, 704/226, 704/231, 704/246, 704/250, 704/270, 704/275, 704/251, 704/252, 704/253, 704/254, 704/257, 704/233, 704/244, 704/247
• US Class Current: 704/244
• CPC Class Codes: G10L 15/20 Speech recognition techniques specially adapted for robustness in adverse environments, e.g. in noise, of stress induced speech G10L21/02 takes precedence

Abstract

A process for voice recognition in a harsh environment such as the cockpit of a military aircraft. An acquisition device is connected to an acoustico-phonetic decoding device which is then connected to a recognition supervisor. The acquisition device is also connected to a voice recognition device which is also connected to the recognition supervisor. The voice recognition device is also connected to a dictionary and to a syntax describing device. The syntax describing device is connected to a dialogue model as well as the recognition supervisor.

Claims

1. A process of voice recognition, comprising the steps of:

performing a coarse recognition of acquired samples, said coarse recognition being based on searching a syntax base,

supplying the N best phrases recognized after comparing results of said coarse recognition with stored acoustic references,

determining phonetic components of said acquired samples by performing an acoustico-phonetic decoding on said acquired samples,

choosing a most appropriate phrase from said N best phrases by comparing said N best phrases with models of probable dialogues and with said phonetic components, and

updating said syntax base as a function of a history of most appropriate phrases chosen in said choosing step.

2. Process according to claim 1, characterized in that the models of dialogue are selected on the basis of the operational context of the recognition.

3. Process according to claim 2, characterized in that the models of dialogue are selected by weighting the syntax arcs between the words, the weighting taking into account, apart from the environmental noise, one at least of the following factors:

the state of sensors of physical quantities relating to the surrounding conditions, the history of the dialogues, the content of a dictionary of acoustic references, the modelling of the dialogue and of the task to which the dialogues relate, the acoustico-phonetic decoding which carries out a first rejection on the basis of phonetic criteria, the various phrases already recognized.

4. Process according to claim 3, applied to a fighter aircraft, characterized in that the physical quantities relating to the surrounding conditions comprise the characteristics of the aircraft'"'"'s navigation and weapons system, and the state of the aircraft'"'"'s sensors including an inertial rig and a load factor.

5. System for voice command in a harsh environment, for implementing the process according to claim 1, characterized in that it includes an acquisition device linked to an acoustico-phonetic decoding device, itself linked to a recognition supervision device, as well as to a voice recognition device, which is itself linked to a dictionary, to the recognition supervisor device and to a syntax recognition device, the latter being linked to a dialogue storage device, which is linked to the supervisor device.

6. System according to claim 5, characterized in that it is linked to a memory card interface, co-operating with customization memory cards in each of which are stored parameters relating to the corresponding speaker.

7. System according to claim 6, characterized in that the voice recognition device is a pattern recognition and signal parameterization device.

8. System according to claim 7, characterized in that the acoustico-phonetic decoder device is an analysis device having banks of filters.

9. System according to claim 6, characterized in that the acoustico-phonetic decoder device is an analysis device having banks of filters.

10. System according to claim 5, characterized in that the voice recognition device is a pattern recognition and signal parametrization device.

11. System according to claim 10, characterized in that the acoustico-phonetic decoder device is an analysis device having banks of filters.

12. System according to claim 5, characterized in that the acoustico-phonetic decoder device is an analysis device having banks of filters.

13. The process of claim 1, further comprising updating said syntax base as a function of environment parameters.

14. The process of claim 13, wherein updating said syntax base is performed as a function of environment parameters corresponding to an aircraft environment.

15. The process of claim 13, wherein updating said syntax base is performed as a function of environment parameters corresponding to a sea vehicle environment.

16. The process of claim 13, wherein updating said syntax base is performed as a function of environment parameters corresponding to a land vehicle environment.

17. The process of claim 13, wherein updating said syntax base is performed as a function of environment parameters corresponding to a metallurgical plant environment.

18. The process of claim 1, further comprising updating said syntax base as a function of aircraft specific commands.

19. The process of claim 1, further comprising updating said syntax base as a function of said phonetic components of said acquired samples.

20. The process of claim 1, wherein updating said syntax base is performed so as to allow an operator to correct an error of recognition or an error attributable to said operator without a command corresponding to said acquired samples being repeated.