Granted Patent B2

US 12,555,575 · App. 17/915,465 · Granted Feb 17, 2026

Inventors: Xu Li (Beijing, CN); Zeming Chen (Beijing, CN)

Assignee: Beijing Baidu Netcom Science Technology Co., Ltd.

G10L15/22G10L15/02G10L15/08G10L2015/088

Abstract

Claims (66)

1 . A wakeup indicator monitoring method, applied to a voice interaction device to be monitored and used for audio testing of the voice interaction device, the method is performed by a wakeup indicator monitoring device and comprises:

acquiring, by the wakeup indicator monitoring device, M pieces of audio data of the voice interaction device to be monitored, wherein M is a positive integer greater than 1;

determining, by the wakeup indicator monitoring device, a first wakeup confidence for each piece of the M pieces of audio data, the first wakeup confidence indicating a probability that audio data contains a first wakeup word for waking up the device to be monitored;

acquiring, by the wakeup indicator monitoring device, a first audio data with a first wakeup confidence in a target zone in the M pieces of audio data, wherein the wakeup confidence indicates that the audio data in the target zone comprises a wakeup word for waking up an audio device; and

determining, by the wakeup indicator monitoring device, the ratio of the first audio data to the M pieces of audio data as a wakeup rate of the device to be monitored, where a wakeup indicator of the device to be monitored comprises the wakeup rate;

wherein prior to obtaining M pieces of audio data for a device to be monitored, the method further comprises:

acquiring, by the wakeup indicator monitoring device, P pieces of audio data of N audio devices and an annotation result of the P pieces of audio data, wherein the annotation result indicates whether the audio data contains a second wakeup word for waking up the audio devices, N is a positive integer, and P is a positive integer greater than 1;

determining, by the wakeup indicator monitoring device, a second wakeup confidence of each piece of the P pieces of audio data; and

counting, by the wakeup indicator monitoring device, an zone where a second wakeup confidence of second audio data of which the ratio is greater than a pre-set threshold value in the P pieces of audio data is located, and obtaining the target zone, wherein the second audio data is the audio data with an annotation result that characterizes it contains the audio data of the second wakeup word;

wherein the P pieces of audio data are obtained from audio log data of the N audio devices, the audio log data comprising a plurality of audio data, and the acquiring P pieces of audio data of N audio devices comprises:

classifying, by the wakeup indicator monitoring device, each piece of audio data in the audio log data by L dimensions respectively so as to obtain L pieces of classification characteristic information about each piece of audio data in the audio log data, L being a positive integer;

determining, by the wakeup indicator monitoring device, audio feature information for each dimension based on the classified feature information of the audio log data;

respectively sampling, by the wakeup indicator monitoring device, in the audio log data based on audio feature information about each dimension so as to obtain an audio sampling result of the L dimensions; and

generating, by the wakeup indicator monitoring device, the P pieces of audio data comprising the audio sampling results of the L dimensions.

2 . The method of claim 1 , wherein the determining a first wakeup confidence for each piece of the M pieces of audio data comprises:

performing feature extraction on target audio data to obtain audio features of the target audio data, wherein the target audio data is any one of the M pieces of audio data; and

scoring the target audio data based on the audio features to obtain a first wakeup confidence of the target audio data.

3 . A non-transitory computer-readable storage medium storing computer instructions for causing a computer to perform the method of claim 1 .

4 . A wakeup indicator monitoring method, applied to a voice interaction device to be monitored and used for audio testing of the voice interaction device, the method is performed by a wakeup indicator monitoring device and comprises:

acquiring, by the wakeup indicator monitoring device, M pieces of audio data of the voice interaction device to be monitored, wherein M is a positive integer greater than 1;

determining, by the wakeup indicator monitoring device, a first wakeup confidence for each piece of the M pieces of audio data, the first wakeup confidence indicating a probability that audio data contains a first wakeup word for waking up the device to be monitored;

acquiring, by the wakeup indicator monitoring device, first audio data with a first wakeup confidence in a target zone in the M pieces of audio data, wherein the wakeup confidence in the target zone characterizes that the audio data does not contain a wakeup word for waking up an audio device; and

determining, by the wakeup indicator monitoring device, the ratio of the first audio data to the M pieces of audio data as a false wakeup rate of the device to be monitored, where a wakeup indicator of the device to be monitored comprises the false wakeup rate;

wherein prior to obtaining M pieces of audio data for a device to be monitored, the method further comprises:

acquiring, by the wakeup indicator monitoring device, P pieces of audio data of N audio devices and an annotation result of the P pieces of audio data, wherein the annotation result indicates whether the audio data contains a second wakeup word for waking up the audio devices, N is a positive integer, and P is a positive integer greater than 1;

determining, by the wakeup indicator monitoring device, a second wakeup confidence of each piece of the P pieces of audio data; and

counting, by the wakeup indicator monitoring device, an zone where a second wakeup confidence of second audio data of which the ratio is greater than a pre-set threshold value in the P pieces of audio data is located, and obtaining the target zone, wherein the second audio data is the audio data with an annotation result that characterizes it contains the audio data of the second wakeup word;

wherein the P pieces of audio data are obtained from audio log data of the N audio devices, the audio log data comprising a plurality of audio data, and the acquiring P pieces of audio data of N audio devices comprises:

classifying, by the wakeup indicator monitoring device, each piece of audio data in the audio log data by L dimensions respectively so as to obtain L pieces of classification characteristic information about each piece of audio data in the audio log data, L being a positive integer;

determining, by the wakeup indicator monitoring device, audio feature information for each dimension based on the classified feature information of the audio log data;

respectively sampling, by the wakeup indicator monitoring device, in the audio log data based on audio feature information about each dimension so as to obtain an audio sampling result of the L dimensions; and

generating, by the wakeup indicator monitoring device, the P pieces of audio data comprising the audio sampling results of the L dimensions.

5 . An electronic device comprising:

at least one processor; and

a memory communicatively coupled to the at least one processor; wherein,

the memory stores instructions are executable by the at least one processor to enable the at least one processor to perform the method of claim 4 .

6 . A non-transitory computer-readable storage medium storing computer instructions for causing a computer to perform the method of claim 4 .

7 . An electronic device comprising:

at least one processor; and

a memory communicatively coupled to the at least one processor; wherein,

the memory stores instructions are executable by the at least one processor to enable the at least one processor to perform a wakeup indicator monitoring method, the method is applied to a voice interaction device to be monitored and used for audio testing of the voice interaction device, the method is performed by a wakeup indicator monitoring device and comprises:

acquiring M pieces of audio data of a voice interaction device to be monitored, wherein M is a positive integer greater than 1;

determining a first wakeup confidence for each piece of the M pieces of audio data, the first wakeup confidence indicating a probability that audio data contains a first wakeup word for waking up the device to be monitored;

acquiring a first audio data with a first wakeup confidence in a target zone in the M pieces of audio data, wherein the wakeup confidence indicates that the audio data in the target zone comprises a wakeup word for waking up an audio device; and

determining the ratio of the first audio data to the M pieces of audio data as a wakeup rate of the device to be monitored, where a wakeup indicator of the device to be monitored comprises the wakeup rate;

wherein prior to obtaining M pieces of audio data for a device to be monitored, the method further comprises:

acquiring P pieces of audio data of N audio devices and an annotation result of the P pieces of audio data, wherein the annotation result indicates whether the audio data contains a second wakeup word for waking up the audio devices, N is a positive integer, and P is a positive integer greater than 1;

determining a second wakeup confidence of each piece of the P pieces of audio data; and

counting an zone where a second wakeup confidence of second audio data of which the ratio is greater than a pre-set threshold value in the P pieces of audio data is located, and obtaining the target zone, wherein the second audio data is the audio data with an annotation result that characterizes it contains the audio data of the second wakeup word;

wherein the P pieces of audio data are obtained from audio log data of the N audio devices, the audio log data comprising a plurality of audio data, and the acquiring P pieces of audio data of N audio devices comprises:

classifying each piece of audio data in the audio log data by L dimensions respectively so as to obtain L pieces of classification characteristic information about each piece of audio data in the audio log data, L being a positive integer;

determining audio feature information for each dimension based on the classified feature information of the audio log data;

respectively sampling in the audio log data based on audio feature information about each dimension so as to obtain an audio sampling result of the L dimensions; and

generating the P pieces of audio data comprising the audio sampling results of the L dimensions.

8 . The electronic device of claim 7 , wherein the at least one processor is configured to perform:

prior to obtaining M pieces of audio data for a device to be monitored, acquiring P pieces of audio data of N audio devices and an annotation result of the P pieces of audio data, wherein the annotation result indicates whether the audio data contains a second wakeup word for waking up the audio devices, N is a positive integer, and P is a positive integer greater than 1;

determining a second wakeup confidence of each piece of the P pieces of audio data; and

counting an zone where a second wakeup confidence of second audio data of which the ratio is greater than a pre-set threshold value in the P pieces of audio data is located, and obtaining the target zone, wherein the second audio data is the audio data with an annotation result that characterizes it contains the audio data of the second wakeup word.

9 . The electronic device of claim 8 , wherein the P pieces of audio data are obtained from audio log data of the N audio devices, the audio log data comprising a plurality of audio data, and the acquiring P pieces of audio data of N audio devices comprises:

classifying each piece of audio data in the audio log data by L dimensions respectively so as to obtain L pieces of classification characteristic information about each piece of audio data in the audio log data, L being a positive integer;

determining audio feature information for each dimension based on the classified feature information of the audio log data;

respectively sampling in the audio log data based on audio feature information about each dimension so as to obtain an audio sampling result of the L dimensions; and

generating the P pieces of audio data comprising the audio sampling results of the L dimensions.

10 . The electronic device of claim 7 , wherein the determining a first wakeup confidence for each piece of the M pieces of audio data comprises:

performing feature extraction on target audio data to obtain audio features of the target audio data, wherein the target audio data is any one of the M pieces of audio data; and

scoring the target audio data based on the audio features to obtain a first wakeup confidence of the target audio data.

Assignments (1)

Priority Claims (1)

Continuity (1)

References Cited (56)