Method and electronic device for processing voice data

• US 10,755,705 B2
• Filed: 11/15/2017
• Issued: 08/25/2020
• Est. Priority Date: 03/29/2017
• Status: Active Grant

• Current Assignee: Lenovo Company Limited (Lenovo Group Ltd.)
• Original Assignee: Lenovo Company Limited (Lenovo Group Ltd.)
• Inventors: Li, Hongwei, Zhang, Dekui
• Primary Examiner(s): Washburn, Daniel C
• Assistant Examiner(s): Ogunbiyi, Oluwadamilola M
• Application Number:
  US15/813,916
• Publication Number:
  US 20180286394A1
• Time in Patent Office: 1,014 Days
• CPC Class Codes: G06F 3/167 Audio in a user interface, e.g. using voice commands for navigating, audio feedback
• CPC Class Codes: G10L 15/00 Speech recognition G10L17/00 takes precedence
• 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 15/30 Distributed recognition, e.g. in client-server systems, for mobile phones or network applications
• CPC Class Codes: G10L 17/24 the user being prompted to utter a password or a predefined phrase
• CPC Class Codes: G10L 2015/223 Execution procedure of a spoken command

Abstract

A data processing method includes acquiring voice data collected by at least two collecting devices from a voice source, acquiring a recognition result of the voice data that corresponds to a voice generated by the voice source, and executing a corresponding command in response to the recognition result.

Claims

1. A processing method, comprising:

acquiring voice data, the voice data being collected by at least two collecting devices from a voice source that generates a sound, each of the at least two collecting devices including a plurality of microphones forming one or more microphone arrays for performing signal processing locally;

calculating a distance between the voice source and each of the at least two collecting devices based on different timings that the sound reaches the plurality of microphones;

stitching the voice data based on a sequence of timings at which the voice data are collected to generate stitched voice data, the stitched voice data including first voice data and second voice data adjacent to each other in the sequence and being collected by different ones of the at least two collecting devices;

analyzing frequencies of the stitched voice data to determine whether a similarity between a first frequency waveform of the first voice data and a second frequency waveform of the second voice data exceeds a threshold;

determining, in response to determining that the similarity exceeds the threshold, that the stitched voice data includes a first content corresponding to the first frequency waveform and a second content corresponding to the second frequency waveform, the first content and the second content being the same as each other and being collected by different ones of the at least two collecting devices during two time periods that overlap with each other;

selecting, according to the calculated distance, one the first content and the second content that is collected by one of the at least two collecting devices closer to the voice source as a target content; and

replacing the first content and the second content with the target content to obtain to-be-recognized voice data for recognition;

acquiring a recognition result of the to-be-recognized voice data, the recognition result corresponding to a voice generated by the voice source; and

in response to the recognition result, executing a corresponding command.

2. The method according to claim 1, wherein stitching the voice data based on the sequence of timings at which the voice data is collected to generate the stitched voice data includes:

determining to-be-stitched collecting devices from the at least two collecting devices based on a voiceprint property of the voice source; and

stitching the voice data collected by the to-be-stitched collecting devices based on a sequence of timings at which the voice data is collected by the to-be-stitched collecting devices.

3. The method according to claim 1, further comprising:

determining one of the at least two collecting devices that is closest to the voice source as a target collecting device; and

determining the voice data collected by the target collecting device as to-be-stitched voice data.

4. The method according to claim 1, further comprising:

determining to-be-stitched collecting devices based on a voiceprint property of the voice source;

identifying the voice data collected by the to-be-stitched collecting devices based on a sequence of timing at which the voice data is collected by the to-be-stitched collecting devices; and

sending the identified voice data to a back-end device for stitching the identified voice data.

5. The method according to claim 1, wherein in response to the recognition result, executing the corresponding command includes:

determining an executing device that matches the voice source as a target executing device; and

sending the corresponding command to the target executing device.

6. The method according to claim 1, wherein in response to the recognition result, executing the corresponding command includes:

determining a target executing device based on a configured operation; and

sending the corresponding command to the target executing device.

7. The method according to claim 1, further comprising:

acquiring a wake-up keyword obtained by analyzing the voice data collected by at least one of the at least two collecting devices that is in a keyword collecting state;

determining at least two to-be-awakened collecting devices according to the wake-up keyword; and

sending an awakening command to the at least two to-be-awakened collecting devices to control the at least two to-be-awakened collecting devices to switch from the keyword collecting state to a command collecting state,

wherein acquiring the voice data includes receiving the voice data collected by the at least two to-be-awakened collecting devices after switching to the command collecting state.

8. The method according to claim 1, wherein the one of the first content and the second content that is selected as the target content has higher strength than another one of the first content and the second content.

9. The method according to claim 1, further comprising:

in response to determining that the stitched voice data does not include the first content and the second content, determining the stitched voice data as the to-be-recognized voice data.

10. The method according to claim 1, wherein:

in a process of collecting the voice data of the voice source by the at least two collecting devices, a relative location of the voice source changes with respect to the at least two collecting devices.

11. An electronic device comprising:

a processor communicatively coupled to at least two collecting devices, each of the at least two collecting devices including a plurality of microphones forming one or more microphone arrays for performing signal processing locally, wherein the processor;

acquires voice data collected by the at least two collecting devices from a voice source,

calculates a distance between the voice source and each of the at least two collecting devices based on different timings that the sound reaches the plurality of microphones;

stitches the voice data based on a sequence of timings at which the voice data are collected to generate stitched voice data, the stitched voice data including first voice data and second voice data adjacent to each other in the sequence and being collected by different ones of the at least two collecting devices,

analyzes frequencies of the stitched voice data to determine whether a similarity between a first frequency waveform of the first voice data and a second frequency waveform of the second voice data exceeds a threshold,

determines, in response to determining that the similarity exceeds the threshold, that the stitched voice data includes a first content corresponding to the first frequency waveform and a second content corresponding to the second frequency waveform, the first content and the second content being the same as each other and being collected by different ones of the at least two collecting devices during two time periods that overlap with each other,

selects, according to the calculated distance, the first content and the second content that is collected by one of the at least two collecting devices closer to the voice source as a target content, and

replaces the first content and the second content with the target content to obtain to-be-recognized voice data for recognition,

acquires a recognition result of the to-be-recognized voice data, the recognition result corresponding to a voice generated by the voice source, and

in response to the recognition result, executes a corresponding command.

12. The electronic device according to claim 11, wherein the processor further:

determines to-be-stitched collecting devices from the at least two collecting devices based on a voiceprint property of the voice source; and

stitches the voice data collected by the to-be-stitched collecting devices based on a sequence of timings at which the voice data is collected by the to-be-stitched collecting devices.

13. The electronic device according to claim 11, wherein the processor further:

determines one of the at least two collecting devices that is closest to the voice source as a target collecting device, and

determines the voice data collected by the target collecting device as to-be-stitched voice data.

14. The electronic device according to claim 11, wherein the processor further:

determines to-be-stitched collecting devices based on a voiceprint property of the voice source;

identifies the voice data collected by each collecting device based on a sequence of timings at which the voice data is collected by the to-be-stitched collecting devices; and

sends the identified voice data to a back-end device for stitching the identified voice data.