Speech recognition system with abbreviated training

• US 10,008,199 B2
• Filed: 08/22/2015
• Issued: 06/26/2018
• Est. Priority Date: 08/22/2015
• Status: Active Grant

• Current Assignee: Toyota Motor Engineering & Manufacturing North America Incorporated (Toyota Motor Corporation)
• Original Assignee: Toyota Motor Engineering & Manufacturing North America Incorporated (Toyota Motor Corporation)
• Inventors: Pierfelice, Jeffrey E., Helm, Sean L., Yamasaki, Bryan E.
• Primary Examiner(s): Riley, Marcus T
• Application Number:
  US14/833,059
• Publication Number:
  US 20170053645A1
• Time in Patent Office: 1,039 Days
• Field of Search: None
• CPC Class Codes: B60R 16/0373 Voice control in general G10L
• CPC Class Codes: G10L 15/07 to the speaker

Abstract

A method of adapting a speech recognition system to its user includes gathering information about a user of a speech recognition system, selecting at least a part of a speech model reflecting estimated speech attributes of the user based on the information about the user, running, in the speech recognition system, a speech model including the selected at least a part of a speech model, and training, in the speech recognition system, other parts of the speech model to reflect identified speech attributes of the user.

Claims

1. A method of adapting a speech recognition system to its user, comprising:

gathering information about a user of a speech recognition system; and

generating a speech model on which the speech recognition system is configured to rely in order to identify speech in utterances of the user, the generating including;

selecting, based on the information about the user, at least a part of a speech model reflecting non-utterance-based estimated speech attributes of the user that differ from those of a greater language speaking population;

running, in the speech recognition system, a speech model including the selected at least a part of a speech model, whereby the speech recognition system relies on the speech model in order to identify speech in the user'"'"'s utterances; and

training, in the speech recognition system, other parts of the speech model, whereby the other parts of the speech model reflect utterance-based identified speech attributes of the user that differ from those of the greater language speaking population;

whereby

the generated speech model includes the selected at least a part of a speech model and the trained other parts of the speech model.

2. The method of claim 1, wherein the information about the user includes any combination of the user'"'"'s accent, age, ethnicity and gender.

3. The method of claim 1, further comprising:

creating a speech model including the selected at least a part of a speech model, wherein the running comprises running, in the speech recognition system, the created speech model, whereby the speech recognition system relies on the created speech model in order to identify speech in the user'"'"'s utterances.

4. The method of claim 1, wherein the selecting and running comprise:

selecting, based on the information about the user, a speech model reflecting the user'"'"'s non-utterance-based estimated speech attributes from a plurality of different speech models respectively reflecting different speech attributes that differ from those of the greater language speaking population; and

running, in the speech recognition system, the selected speech model, whereby the speech recognition system relies on the selected speech model in order to identify speech in the user'"'"'s utterances.

5. The method of claim 1, wherein the gathering and selecting comprise:

receiving, at the speech recognition system, the information about the user; and

selecting, based on the information about the user, with the speech recognition system, the at least a part of a speech model from a plurality of different at least parts of speech models included in the speech recognition system and respectively reflecting different speech attributes that differ from those of the greater language speaking population.

6. The method of claim 1, wherein the gathering and selecting comprise:

receiving the information about the user;

selecting, based on the information about the user, the at least a part of a speech model from a plurality of different at least parts of speech models included in a speech server remote from the speech recognition system and respectively reflecting different speech attributes that differ from those of the greater language speaking population; and

transferring, to the speech recognition system from the speech server, the selected at least a part of a speech model.

7. The method of claim 1, further comprising:

receiving, at the speech recognition system, an utterance from the user; and

identifying, with the speech recognition system, the user'"'"'s utterance-based speech attributes based on the received utterance.

8. A method of adapting systems of a vehicle, comprising:

gathering information about a user of a vehicle including a speech recognition system and at least one other system;

generating a speech model on which the speech recognition system is configured to rely in order to identify speech in utterances of the user, the generating including;

selecting, based on the information about the user, at least a part of a speech model reflecting non-utterance-based estimated speech attributes of the user that differ from those of a greater language speaking population;

running, in the speech recognition system, a speech model including the selected at least a part of a speech model, whereby the speech recognition system relies on the speech model in order to identify speech in the user'"'"'s utterances; and

training, in the speech recognition system, other parts of the speech model, whereby the other parts of the speech model reflect utterance-based identified speech attributes of the user that differ from those of the greater language speaking population;

whereby

the generated speech model includes the selected at least a part of a speech model and the trained other parts of the speech model; and

adapting, based on the information about the user, the at least one other system.

9. The method of claim 8, wherein the information about the user includes any combination of the user'"'"'s accent, age, ethnicity and gender.

10. The method of claim 8, wherein the information about the user includes a location where the user acquired the vehicle, and the gathering comprises:

determining the vehicle'"'"'s location upon the speech recognition system'"'"'s initial startup; and

identifying the determined location as the location where the user acquired the vehicle.

11. The method of claim 8, wherein the vehicle includes input hardware, and the gathering comprises:

receiving the information about the user at the input hardware.

12. The method of claim 8, wherein the gathering comprises:

receiving the information about the user from among information documenting the user'"'"'s acquisition of the vehicle.

13. The method of claim 8, further comprising:

creating a speech model including the selected at least a part of a speech model, wherein the running comprises running, in the speech recognition system, the created speech model, whereby the speech recognition system relies on the created speech model in order to identify speech in the user'"'"'s utterances.

14. The method of claim 8, wherein the selecting and running comprise:

selecting, based on the information about the user, a speech model reflecting the user'"'"'s non-utterance-based estimated speech attributes from a plurality of different speech models respectively reflecting different speech attributes that differ from those of the greater language speaking population; and

running, in the speech recognition system, the selected speech model, whereby the speech recognition system relies on the selected speech model in order to identify speech in the user'"'"'s utterances.

15. The method of claim 8, wherein the gathering and selecting comprise:

receiving, at the speech recognition system, the information about the user; and

selecting, based on the information about the user, with the speech recognition system, the at least a part of a speech model from a plurality of different at least parts of speech models included in the speech recognition system and respectively reflecting different speech attributes that differ from those of the greater language speaking population.

16. The method of claim 8, wherein the gathering and selecting comprise:

receiving the information about the user;

selecting, based on the information about the user, the at least a part of a speech model from a plurality of different at least parts of speech models included in a speech server remote from the speech recognition system and respectively reflecting different speech attributes that differ from those of the greater language speaking population; and

transferring, to the speech recognition system from the speech server, the selected at least a part of a speech model.

17. The method of claim 8, further comprising:

receiving, at the speech recognition system, an utterance from the user; and

identifying, with the speech recognition system, the user'"'"'s utterance-based speech attributes based on the received utterance.

18. The method of claim 8, wherein the at least one other system is an IVI system supported by the speech recognition system that has radio presets settable to different radio stations, the information about the user includes a location where the user acquired the vehicle, and the adapting comprises:

setting one or more of the radio presets to radio stations local to the location where the user acquired the vehicle.

19. The method of claim 8, wherein the at least one other system is an engine system that has a plurality of engine modes, the information about the user includes a location where the user acquired the vehicle, and the adapting comprises:

selecting an engine mode from the plurality of engine modes based on the location where the user acquired the vehicle; and

setting the engine system to the selected engine mode.

20. The method of claim 8, wherein the at least one other system is a suspension system that has a plurality of suspension modes, and the adapting comprises:

selecting a suspension mode from the plurality of suspension modes based on the information about the user; and

setting the suspension system to the selected suspension mode.