Utterance processing for network-based speech recognition utilizing a client-side cache

• US 8,224,644 B2
• Filed: 12/18/2008
• Issued: 07/17/2012
• Est. Priority Date: 12/18/2008
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Krumel, Andrew K., Chang, Shuangyu, Chambers, Robert L.
• Primary Examiner(s): MCFADDEN, SUSAN IRIS
• Application Number:
  US12/337,810
• Publication Number:
  US 20100161328A1
• Time in Patent Office: 1,307 Days
• Field of Search: 704/236
• US Class Current: 704/236
• CPC Class Codes: G10L 15/285 Memory allocation or algorithm optimisation to reduce hardware requirements
• CPC Class Codes: G10L 15/30 Distributed recognition, e.g. in client-server systems, for mobile phones or network applications
• CPC Class Codes: G10L 2015/227 of the speaker; Human-factor methodology

Abstract

Embodiments are provided for utilizing a client-side cache for utterance processing to facilitate network based speech recognition. An utterance comprising a query is received in a client computing device. The query is sent from the client to a network server for results processing. The utterance is processed to determine a speech profile. A cache lookup is performed based on the speech profile to determine whether results data for the query is stored in the cache. If the results data is stored in the cache, then a query is sent to cancel the results processing on the network server and the cached results data is displayed on the client computing device.

Claims

1. A computer-implemented method for utilizing a client-side cache for utterance processing to facilitate network based speech recognition, comprising:

receiving an utterance comprising a query in the client;

sending the query from the client to a network server for results processing;

processing the utterance to determine a profile at the client, wherein processing the utterance to determine a profile at the client comprises;

dividing the utterance into a plurality of frames comprising a time series of numerical sequences, the plurality of frames comprising the duration of the utterance; and

extracting at least one speech component from the utterance;

performing a lookup based on the profile to determine whether results data for the query is stored in the cache, wherein performing a lookup based on the profile to determine whether results data for the query is stored in the cache comprises;

aligning the plurality of frames in the utterance with a plurality of frames comprising at least one stored utterance in the cache;

extracting at least one speech component from the at least one stored utterance in the cache;

calculating a score based on at least one of;

a degree of the alignment of the plurality of frames in the utterance with the plurality of frames comprising the at least one stored utterance in the cache; and

a similarity between the at least one speech component extracted from the utterance and the at least one speech component extracted from the at least one stored utterance in the cache; and

determining whether the results data for the query is stored in the cache based on the score; and

if the results data is stored in the cache, then displaying the cached results data on the client.

2. The method of claim 1 further comprising, if the results data is stored in the cache, then sending a query to cancel the results processing on the network server.

3. The method of claim 1 further comprising, if the results data is not stored in the cache, then:

receiving the results data from the network server; and

displaying the results data from the network server on the client.

4. The method of claim 1, wherein the steps of sending the query from the client to a network server for results processing and processing the utterance to determine a profile at the client occur nearly simultaneously.

5. The method of claim 1, wherein receiving an utterance comprising a query in the client comprises receiving the utterance in an audio capture device.

6. The method of claim 1, wherein extracting at least one speech component from the utterance comprises extracting at least one of a syllable rate, a pitch, a syllable count, and a stress pattern from the utterance and wherein extracting at least one speech component from the at least one stored utterance in the cache comprises extracting at least one of a syllable rate, a pitch, a syllable count, and a stress pattern from the at least one stored utterance.

7. A system for utilizing a client-side cache for utterance processing to facilitate network based speech recognition, the system comprising a client computer, the client computer comprising:

a memory for storing executable program code; and

a processor, functionally coupled to the memory, the processor being responsive to computer-executable instructions contained in the program code and operative;

to receive an utterance comprising a query;

to send the query to a network server for results processing;

to process the utterance to determine a profile by;

dividing the utterance into a plurality of frames comprising a time series of numerical sequences, the plurality of frames comprising the duration of the utterance; and

extracting at least one speech component from the utterance;

to perform a lookup based on the profile to determine whether results data for the query is stored in the cache; and

if the results data is stored in the cache, then to display the cached results data.

8. The system of claim 7, wherein the processor, in performing a lookup based on the profile to determine whether results data for the query is stored in the cache, is further operative to:

to align the plurality of frames in the utterance with a plurality of frames comprising at least one stored utterance in the cache;

to extract at least one speech component from the at least one stored utterance in the cache;

to calculate a score based on at least one of;

a degree of the alignment of the plurality of frames in the utterance with the plurality of frames comprising the at least one stored utterance in the cache; and

a similarity between the at least one speech component extracted from the received utterance and the at least one speech component extracted from the at least one stored utterance in the cache; and

to determine whether the results data for the query is stored in the cache based on the score.

9. The system of claim 8, wherein the at least one speech component extracted from the utterance comprises at least one of a syllable rate, a pitch, a syllable count, and a stress pattern and wherein the at least one speech component extracted from the at least one stored utterance in the cache comprises at least one of a syllable rate, a pitch, a syllable count, and a stress pattern.

10. The system of claim 7, wherein if the results data is stored in the cache, then the processor is further operative to send a query to cancel the results processing on the network server.

11. The system of claim 7, wherein if the results data is not stored in the cache then the processor is further operative to:

receive the results data from the network server; and

display the results data from the network server on the client.

12. The system of claim 7, wherein the query is sent to the network server for results processing and the utterance is processed to determine the profile nearly simultaneously.

13. A computer-readable storage device comprising computer executable instructions which, when executed on a computer, will cause the computer to perform a method of utilizing a client-side cache for utterance processing to facilitate network based speech recognition, the method comprising:

receiving an utterance comprising a query;

sending the query from the client to a network server for results processing;

processing the utterance to determine a profile at the client, wherein processing the utterance to determine a profile at the client comprises;

dividing the utterance into a plurality of frames comprising a time series of numerical sequences, the plurality of frames comprising the duration of the utterance; and

extracting at least one speech component from the utterance, wherein the steps of sending the query from the client to a network server for results processing and processing the utterance to determine a profile at the client occur nearly simultaneously;

performing a lookup based on the profile to determine whether results data for the query is stored in the cache;

if the results data is stored in the cache, then

sending a query to cancel the results processing on the network server; and

displaying the cached results data on the client;

if the results data is not stored in the cache, then

receiving the results data from the network server; and

displaying the results data from the network server on the client.

14. The computer-readable storage device of claim 13, wherein performing a lookup based on the profile to determine whether results data for the query is stored in the cache comprises:

aligning the plurality of frames in the utterance with a plurality of frames comprising at least one stored utterance in the cache;

extracting at least one speech component from the at least one stored utterance in the cache;

calculating a score based on at least one of;

a degree of the alignment of the plurality of frames in the utterance with the plurality of frames comprising the at least one stored utterance in the cache; and

a similarity between the at least one speech component extracted from the utterance and the at least one speech component extracted from the at least one stored utterance in the cache; and

determining whether the results data for the query is stored in the cache based on the score.

15. The computer-readable storage device of claim 14, wherein extracting at least one speech component from the utterance comprises extracting at least one of a syllable rate, a pitch, a syllable count, and a stress pattern from the utterance.

16. The computer-readable storage device of claim 14, wherein extracting at least one speech component from the at least one stored utterance in the cache comprises extracting at least one of a syllable rate, a pitch, a syllable count, and a stress pattern from the at least one stored utterance.