System and method for improving recognition accuracy in speech recognition applications

• US 7,457,751 B2
• Filed: 11/30/2004
• Issued: 11/25/2008
• Est. Priority Date: 11/30/2004
• Status: Active Grant

• Current Assignee: Stryker Corporation
• Original Assignee: Vocera Communications, Inc. (Stryker Corporation)
• Inventors: Shostak, Robert E.
• Primary Examiner(s): Opsasnick; Michael N
• Application Number:
  US11/000,590
• Publication Number:
  US 20060116885A1
• Time in Patent Office: 1,456 Days
• Field of Search: 455/569.1, 455/563, 704/251, 704/254
• US Class Current: 704/251
• CPC Class Codes: G10L 15/18 using natural language modelling
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue
• CPC Class Codes: G10L 2015/228 of application context

Abstract

A speech recognition system and method are provided to correctly distinguish among multiple interpretations of an utterance. This system is particularly useful when the set of possible interpretations is large, changes dynamically, and/or contains items that are not phonetically distinctive. The speech recognition system extends the capabilities of mobile wireless communication devices that are voice operated after their initial activation.

Claims

1. A wireless communications system, comprising:

a controlling computer;

one or more wireless access points connected to the controlling computer by a network;

a badge that communicates using a wireless protocol with one of the wireless access points and receives a voice command; and

wherein the controlling computer further comprises a speech recognition system that receives a voice command from a particular user through the badge and interprets the voice command of the user to generate a set of resulting voice command interpretations, the speech recognition system further comprising a database containing a set of names and a set of verbs that are valid for a voice command and a say or spell mechanism that generates a set of grammar wherein the set of grammar further comprises each verb together with each name, a spelling of each verb together with each name, each verb together with a spelling of each name and a spelling of each verb together with a spelling of each name; and

wherein the say and or spell mechanism reduces the chance of an incorrect voice command interpretation.

2. The system of claim 1, wherein each voice command further comprises a verb portion and an argument portion and wherein the say or spell mechanism further comprises a grammar store, the grammar store having for each verb of each voice command, a spelling of the verb of the voice command and for each argument of each voice command, a spelling of the argument so that the grammar store contains the combination of each verb and each argument for the voice commands.

3. The system of claim 2, wherein the speech recognition system further comprises a spelling mechanism that permits the user to spell a voice command verb and argument using the grammar store of the say or spell mechanism.

4. A method for improving the accuracy of a computer-implemented speech recognition system, the method comprising:

obtaining a set of voice command interpretations from a speech recognition engine, each voice command interpretation having a verb portion;

partitioning the set of voice command interpretations into one or more isomorphic sets, each isomorphic set containing one or more voice command interpretations having a common verb portion;

filtering the one or more voice command interpretations in each isomorphic set to generate a preferred voice command interpretation; and

outputting the preferred voice command interpretation.

5. The method of claim 4, wherein the voice command interpretation further comprises an argument portion.

6. The method of claim 5, wherein the filtering the one or more voice command interpretations in each isomorphic set further comprises comparing the argument of each voice command interpretation in each isomorphic set to a store of arguments and removing a particular voice command interpretation when the argument of the particular voice command interpretation does not match an entity in the store of arguments, unless none of the interpretations match the inner circle store, in which case none of the interpretations are removed.

7. The method of claim 6, wherein the store of arguments further comprises an entity store for a particular user from which a voice command has been issued, the store of entities further comprises a name of one or more of an individual person, a group of people, an organization and a group of organizations.

8. The method of claim 7 further comprising generating an entity store for each user.

9. The method of claim 8, wherein generating the entity store further comprises one or more of manually setting an entity store for a user, setting an entity store of the user based on a buddy list of the user, setting an entity store of the user based on the department of the user and automatically setting an entity store for the user based on the arguments of prior voice commands issued by the user.

10. The method of claim 8 further comprising removing an entity from the entity store for the user.

11. The method of claim 10, wherein removing an entity from the entity store further comprises one or more of manually removing an entity from the entity store and automatically removing an entity from the entity store when the entity has not appeared as an argument for a voice command of the user for a predetermined time period.

12. The method of claim 5, wherein obtaining the set of voice command interpretations further comprises spelling the verb portion of the voice command to distinguish similar voice command interpretations and generate a set of voice command interpretations and spelling the argument portion of the voice command to further distinguish similar voice command interpretations to generate a reduced set of voice command interpretations.

13. The method of claim 12, wherein obtaining the set of voice command interpretations further comprises storing a grammar store wherein the grammar store further comprises, for each verb of each voice command, a spelling of the verb of the voice command and for each argument of each voice command, a spelling of the argument so that the grammar store contains the combination of each verb and each argument for the voice commands.

14. The method of claim 13, wherein the spelling steps further comprises comparing a spelled voice command to the grammar store.

15. A method for improving the accuracy of a computer-implemented speech recognition system, the method comprising:

receiving a voice command from a user, the voice command having a verb portion and an argument portion;

spelling the verb portion of the voice command to distinguish similar voice command interpretations and generate a set of voice command interpretations; and

spelling the argument portion of the voice command to further distinguish similar voice command interpretations to generate a reduced set of voice command interpretations.

16. The method of claim 15 further comprising storing a grammar store wherein the grammar store further comprises, for each verb of each voice command, a spelling of the verb of the voice command and for each argument of each voice command, a spelling of the argument so that the grammar store contains the combination of each verb and each argument for the voice commands.

17. The method of claim 16, wherein the spelling steps further comprises comparing a spelled voice command to the grammar store.

18. A speech recognition system, comprising:

a speech recognition engine that generates a set of voice command interpretations based on a voice command of a user, the voice command having a verb portion and an argument portion;

the speech recognition engine further comprises a database containing a set of names and a set of verbs that are valid for a voice command and a say or spell mechanism that generates a set of grammar wherein the set of grammar further comprises each verb together with each name, a spelling of each verb together with each name, each verb together with a spelling of each name and a spelling of each verb together with a spelling of each name; and

wherein the say and or spell mechanism reduces the chance of incorrect voice command interpretations.

19. The system of claim 18, wherein the speech recognition system further comprises a spelling mechanism that permits the user to spell a voice command verb and argument using the grammar store of the say or spell mechanism.