Apparatus and method for recognizing voice command

• US 9,142,212 B2
• Filed: 04/26/2011
• Issued: 09/22/2015
• Est. Priority Date: 08/03/2010
• Status: Expired due to Fees

• Current Assignee: Samsung Electronics Co. Ltd.
• Original Assignee: Byung-Kwan Kwak, Chi-Youn Park, Jeong-Mi Cho
• Inventors: Park, Chi-Youn, Kwak, Byung-Kwan, Kim, Jeong-Su, Cho, Jeong-Mi
• Primary Examiner(s): Desir, Pierre-Louis
• Assistant Examiner(s): Sharma, Neeraj
• Application Number:
  US13/093,919
• Publication Number:
  US 20120035935A1
• Time in Patent Office: 1,610 Days
• Field of Search: 704/239, 704/275, 704/215, 704/272, 704/257, 704/9, 704/238, 704/251, 704/246, 704/243, 704/201, 705/10, 379/88.03, 379/88.04, 379/201.01, 379/88.01, 455/566
• US Class Current: 1/1
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue

Abstract

An apparatus and method for recognizing a voice command for use in an interactive voice user interface are provided. The apparatus includes a command intention belief generation unit that is configured to recognize a first voice command and that may generate one or more command intention beliefs for the first voice command. The apparatus also includes a command intention belief update unit that is configured to update each of the command intention beliefs based on a system response to the first voice command and a second voice commands. The apparatus also includes a command intention belief selection unit that is configured to select one of the updated command intention beliefs for the first voice command. The apparatus also includes an operation signal output unit that is configured to select a final command intention from the selected updated command intention belief and to output an operation signal based on the selected final command intention.

Claims

1. An apparatus for recognizing a voice command, the apparatus, comprising:

a probability distribution of a command intention generator, implemented by a processor, configured to recognize a first voice command, and in response, configured to generate probability distributions of the command intention in consideration of predefined possible cases;

a probability distribution of the command intention updater, implemented by a processor, configured to update each of the probability distributions of the command intention based on a system response based on one or the generated probability distributions of the command intention to the first voice command and a second voice command issued in response to the system response;

a probability distribution of the command intention selector, implemented by a processor, configured to select one of the updated probability distributions of the command intention for the first voice command; and

an operation signal generator, implemented by a processor, configured to select a final command intention from the selected updated probability distributions of the command intention and output an operation signal based on the selected final command intention;

wherein the probability distribution of the command intention generator, implemented by the processor therefor, is further configured to generate different probability distributions of the command intention for the first voice command for different possible cases, the probability distribution of the command intention updater, implemented by the processor therefor, is further configured to hierarchically update each of the different probability distributions of the command intention, and the different possible cases include when the intention of the user is uniformly maintained, when the intention of the user is slightly modified, and when the intention of the user is completely changed.

2. The apparatus of claim 1, wherein the probability distribution of the command intention updater comprises:

a second voice command recognizer configured to recognize the intention of the second voice command;

a command intention change detector configured to compare the system response to the first voice command to the intention of the second voice command and determine whether the intention of a user has changed based on the results of the comparison;

a probability distribution of the command intention resetter configured to reset the probability distribution of the command intention for the first voice command based on the results determined by the command intention change detector; and

a command intention updater configured to update the reset probability distributions of the command intention for the first voice command.

3. The apparatus of claim 1, wherein each of the probability distributions of the command intention for the first voice command includes at least one of the intention of the first voice command, information indicating whether the first voice command has been confirmed, information indicating whether the system response to the first voice command has been output, and information indicating whether the output signal has been output.

4. The apparatus of claim 1, wherein the probability distribution of the command intention updater is further configured to continue updating the probability distributions of the command intention for the first voice command until the operation signal output unit selects the final command intention.

5. The apparatus of claim 1, wherein the probability distribution of the command intention selector is further configured to select a lowest-entropy probability distribution of the command intention from the probability distributions of the command intention for the first voice command.

6. The apparatus of claim 3, wherein the probability distribution of the command intention selector is further configured to select one of the updated probability distributions of the command intention for the first voice command based on particular information values of the corresponding probability distributions of the command intention.

7. The apparatus of claim 3, wherein the probability distribution of the command intention selector is further configured to select one of the updated probability distributions of the command intention for the first voice command based on a weighted average of the particular information values of the corresponding probability distributions of the command intention.

8. The apparatus of claim 1, wherein the operation signal generator is further configured to select a final command intention from a highest-probability probability distribution of the command intention selected from the updated probability distributions of the command intention for the first voice command and output an operation signal based on the selected final command intention.

9. The apparatus of claim 1, wherein the operation signal generator is further configured to apply a positive reward value to operation signals that comply with the selected final command intention and a negative reward value to operation signals that do not comply with the selected final command intention, and the apparatus selects the operation signal with a highest reward value as an optimal operation signal for the first voice command.

10. A method of recognizing a voice command, the method comprising:

recognizing a first voice command and generating probability distributions of the command intention for the first voice command in consideration of predefined possible cases;

updating each of the probability distributions of the command intention based on a system response based on one of the generated probability distributions of the command intention to the first voice command and a second voice command issued in response to the system response;

selecting an updated probability distribution of the command intention for the first voice command; and

selecting a final command intention from the selected updated probability distributions of the command intention outputting an operation signal based on the selected final command intention, wherein the method of recognizing the voice command is implemented by at least a processor;

wherein the generating of the probability distributions of the command intention for the first voice command comprises generating different probability distributions of the command intention for the first voice command for different possible cases, the updating of the probability distributions of the command intention for the first voice command comprises hierarchically updating each of the different probability distributions of the command intention, and the different possible cases include when the intention of the user is uniformly maintained, slightly modified, and completely changed.

11. The method of claim 10, wherein the updating of the probability distribution of the command intention for the first voice command comprises:

recognizing the intention of the second voice command;

comparing the system response to the first voice command with the intention of the second voice command and determining whether the intention of a user has changed based on the results of the comparison;

resetting the probability distributions of the command intention for the first voice command based on the results of whether the intention of the user has changed; and

updating the reset probability distributions of the command intention for the first voice command.

12. The method of claim 10, wherein each of the probability distributions of the command intention for the first voice command includes at least one of the intention of the first voice command, information indicating whether the first voice command has been confirmed, information indicating whether the system response to the first voice command has been output, and information indicating whether the output signal has been output.

13. The method of claim 10, wherein the updating of the probability distributions of the command intention for the first voice command comprises continually updating the probability distributions of the command intention for the first voice command until the final command intention is selected.

14. The method of claim 10, wherein the selecting of the updated probability distributions of the command intention for the first voice command comprises selecting a lowest-entropy probability distribution of the command intention from the updated probability distributions of the command intention for the first voice command.

15. The method of claim 12, wherein the selecting of the updated probability distributions of the command intention for the first voice command further comprises selecting one of the updated probability distributions of the command intention for the first voice command based on particular information values of the corresponding probability distributions of the command intention.

16. The method of claim 12, wherein the selecting of the updated probability distribution of the command intention for the first voice command further comprises selecting one of the updated probability distributions of the command intention for the first voice command based on a weighted average of the particular information values of the corresponding probability distributions of the command intention.

17. The method of claim 10, wherein the outputting of the output signal comprises selecting a final command intention from a highest-probability probability distribution of the command intention selected from the updated probability distributions of the command intention for the first voice command and outputting an operation signal based on the selected final command intention.

18. The method of claim 10, wherein the outputting of the output signal comprises applying a positive reward value to operation signals that comply with the selected final command intention and a negative reward value to operation signals that do not comply with the selected final command intention, and selecting one of the operation signals with a highest reward value as an optimal operation signal for the first voice command.