Sound-based call-quality detector

• US 10,244,104 B1
• Filed: 06/14/2018
• Issued: 03/26/2019
• Est. Priority Date: 06/14/2018
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Inventors: Ireland, Richard Clive, Gilbert, Michael T., Clark, Bradford R., Barclay, Thomas Donald
• Primary Examiner(s): Tieu, Binh Kien
• Application Number:
  US16/008,969
• Time in Patent Office: 285 Days
• Field of Search: 379 101, 379 9, 379 1401, 379 1001, 379 12, 379 1501, 379 1505, 379 22, 379 2202, 379 2203, 379 24, 379 2601, 379 2701, 379 2702, 379 2703, 379 2708, 379 2909, 379 291, 379 3201, 379 3204
• CPC Class Codes: G06N 3/084 Backpropagation, e.g. using gradient descent
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue
• CPC Class Codes: G10L 15/26 Speech to text systems G10L15/08 takes precedence
• CPC Class Codes: G10L 25/30 using neural networks
• CPC Class Codes: G10L 25/60 for measuring the quality of voice signals
• CPC Class Codes: H04M 2201/18 Comparators
• CPC Class Codes: H04M 2201/40 using speech recognition speech recognition per se G10L15/00
• CPC Class Codes: H04M 2203/2038 Call context notifications
• CPC Class Codes: H04M 3/2236 Quality of speech transmission monitoring
• CPC Class Codes: H04M 3/2263 Network management

Abstract

System and techniques for sound-based call-quality detector are described herein. An audio sample can be received via a hardware interface from an on-going communication. The audio sample can be processed to produce a transcript. The transcript can be compared to a call-quality dataset to produce a call-quality degradation indication. Here, the call quality dataset includes indicia of communication quality. Network maintenance can then be performed in response to the call-quality degradation indication. The network maintenance can include adjusting a network device setting.

Claims

1. A system for a sound-based call-quality detector, the system comprising:

processing circuitry; and

memory including instructions that, when executed by the processing circuitry, cause the processing circuitry to;

receive, via a hardware interface, an audio sample from an on-going communication;

process the audio sample produce a transcript;

compare the transcript to a call-quality dataset to produce a call-quality degradation indication, the call quality dataset including indicia of communication quality; and

preform network maintenance in response to the call-quality degradation indication, the network maintenance including adjusting a network device setting.

2. The system of claim 1, wherein the transcript is represented within an artificial neural network, wherein, to compare the transcript to the call-quality dataset, the processing circuitry uses the artificial neural network to produce the call-quality degradation indication as output of the artificial neural network.

3. The system of claim 1, wherein the transcript includes a confidence value for words in the transcript.

4. The system of claim 3, wherein, to compare the transcript to the call-quality dataset, the processing circuitry is to:

calculate a variance between the confidence values in the output and historical confidence values in the call-quality dataset; and

produce the call-quality degradation indication in response to the variance surpassing a threshold.

5. The system of claim 1, wherein, to compare the transcript to the call-quality dataset, the processing circuitry is to:

compare phrases from the transcript with phrases in the call-quality dataset; and

produce the call-quality degradation indication in response to matching a phrase from the transcript to a phrase in the library.

6. The system of claim 5, wherein the transcript is supplemented by user-interface interactions of a user.

7. A method for a sound-based call-quality detector, the method comprising:

receiving, via a hardware interface, an audio sample from an on-going communication;

processing the audio sample produce a transcript;

comparing the transcript to a call-quality dataset to produce a call-quality degradation indication, the call quality dataset including indicia of communication quality; and

performing network maintenance in response to the call-quality degradation indication, the network maintenance including adjusting a network device setting.

8. The method of claim 7, wherein the transcript is represented within an artificial neural network, wherein comparing the transcript to the call-quality dataset is performed by the artificial neural network to produce the call-quality degradation indication as output of the artificial neural network.

9. The method of claim 7, wherein the transcript includes a confidence value for words in the transcript.

10. The method of claim 9, wherein comparing the transcript to the call-quality dataset includes:

calculating a variance between the confidence values in the output and historical confidence values in the call-quality dataset; and

producing the call-quality degradation indication in response to the variance surpassing a threshold.

11. The method of claim 7, wherein comparing the transcript to the call-quality dataset includes:

comparing phrases from the transcript with phrases in the call-quality dataset; and

producing the call-quality degradation indication in response to matching a phrase from the transcript to a phrase in the library.

12. The method of claim 11, wherein the transcript is supplemented by user-interface interactions of a user.

13. A device for a sound-based call-quality detector, the device comprising:

means for receiving, via a hardware interface, an audio sample from an on-going communication;

means for processing the audio sample produce a transcript;

means for comparing the transcript to a call-quality dataset to produce a call-quality degradation indication, the call quality dataset including indicia of communication quality; and

means for performing network maintenance in response to the call-quality degradation indication, the network maintenance including adjusting a network device setting.

14. The device of claim 13, wherein the transcript is represented within an artificial neural network, wherein the means for comparing the transcript to the call-quality dataset are performed by the artificial neural network to produce the call-quality degradation indication as output of the artificial neural network.

15. The device of claim 13, wherein the transcript includes a confidence value for words in the transcript.

16. The device of claim 15, wherein the means for comparing the transcript to the call-quality dataset include:

means for calculating a variance between the confidence values in the output and historical confidence values in the call-quality dataset; and

means for producing the call-quality degradation indication in response to the variance surpassing a threshold.

17. The device of claim 13, wherein the means for comparing the transcript to the call-quality dataset include:

means for comparing phrases from the transcript with phrases in the call-quality dataset; and

means for producing the call-quality degradation indication in response to matching a phrase from the transcript to a phrase in the library.

18. The device of claim 17, wherein the transcript is supplemented by user-interface interactions of a user.