Control method and control apparatus for speech interaction

Patent metadata

Field	Value
Publication number	US-11615784-B2
Application number	US-202017118869-A
Country	US
Kind code	B2
Filing date	Dec 11, 2020
Priority date	Jun 30, 2020
Publication date	Mar 28, 2023
Grant date	Mar 28, 2023

Abstract

The present disclosure discloses a control method and a control apparatus for speech interaction. The detailed implementation solution of the control method for the speech interaction includes: collecting an audio signal; detecting a wake-up word in the audio signal to obtain a wake-up word result; and playing a prompt tone and/or executing a speech instruction in the audio signal based on the wake-up word result.

First claim

What is claimed is:

1. A control method for speech interaction, comprising:
collecting an audio signal;
detecting a wake-up word in the audio signal to obtain a wake-up word result; and
playing a prompt tone and/or executing a speech instruction in the audio signal based on the wake-up word result,
wherein the wake-up word result comprises a first confidence and a second confidence, the first confidence is configured to represent a reliability that the audio signal comprises a target wake-up word, the second confidence is configured to represent a reliability that the audio signal comprises an ordinary wake-up word, the number of syllables of the target wake-up word is no more than that of the ordinary wake-up word, and
playing the prompt tone and/or executing the speech instruction in the audio signal based on the wake-up word result comprises:
executing the speech instruction in a case that the first confidence reaches a first confidence threshold;
playing the prompt tone in a case that the second confidence reaches a second confidence threshold and the first confidence fails to reach the first confidence threshold.

2. The method of claim 1 , wherein before or when executing the speech instruction in the audio signal based on the wake-up word result, the method further comprises:
withholding from playing the prompt tone.

3. The method of claim 1 , wherein the ordinary wake-up word comprises at least one target wake-up word; and
detecting the wake-up word in the audio signal comprises:
performing a primary detection on the target wake-up word in the audio signal by employing a wake-up word detection model to obtain a first detection result;
performing a secondary detection on the target wake-up word within a set period after the primary detection to obtain a second detection result; and
determining the first confidence and the second confidence based on the first detection result and the second detection result.

4. The method of claim 1 , wherein the speech instruction is obtained by detecting a part subsequent to the wake-up word in the audio signal.

5. The method of claim 1 , wherein the method is executed by a speech interaction terminal; and
executing the speech instruction in the case that the first confidence reaches the first confidence threshold comprises:
sending the audio signal comprising the target wake-up word and the speech instruction subsequent to the target wake-up word to a server in the case that the first confidence reaches the first confidence threshold, such that the server detects the wake-up word at a front part of the audio signal and the speech instruction subsequent to the wake-up word; and
obtaining the speech instruction from the server and executing the speech instruction.

6. The method of claim 1 , wherein the target wake-up word is a word with less than four syllables; and the ordinary wake-up word is a word with four or more syllables.

7. The method of claim 1 , wherein the number of syllables of the target wake-up word is same as that of the ordinary wake-up word.

8. A method for controlling a speech interaction, comprising:
obtaining an audio signal;
detecting a wake-up word at a front part of the audio signal and detecting a speech instruction subsequent to the wake-up word, to obtain a wake-up word result and a speech instruction result; and
controlling a speech interaction terminal to play a prompt tone and/or to execute the speech instruction based on at least one of the wake-up word result and the speech instruction result,
wherein the wake-up word result comprises a third confidence and a fourth confidence, the third confidence is configured to represent a reliability that the front part of the audio signal comprises a target wake-up word, the fourth confidence is configured to represent a reliability that the front part of the audio signal comprises an ordinary wake-up word, the number of syllables of the target wake-up word is no more than that of the ordinary wake-up word, and
controlling the speech interaction terminal to play the prompt tone and/or to execute the speech instruction based on the at least one of the wake-up word result and the speech instruction result comprises:
controlling the speech interaction terminal to execute the speech instruction based on the speech instruction result in a case that the third confidence reaches a third confidence threshold;
controlling the speech interaction terminal to play the prompt tone in a case that the third confidence fails to reach the third confidence threshold;
controlling the speech interaction terminal to execute the speech instruction and/or to play the prompt tone based on the speech instruction result in a case that the fourth confidence reaches a fourth confidence threshold; and
controlling the speech interaction terminal to send a dummy instruction in a case that the fourth confidence fails to reach the fourth confidence threshold and the third confidence fails to reach the third confidence threshold.

9. The method of claim 8 , wherein detecting the wake-up word at the front part of the audio signal and detecting the speech instruction subsequent to the wake-up word, to obtain the wake-up word result and the speech instruction result comprise:
performing wake-up word detection on a front part of a recognition text of the audio signal to obtain a wake-up word detection result of the front part;
determining an interaction confidence of the audio signal based on at least one of an acoustic feature representation of the audio signal and a textual feature representation associated with the recognition text of the audio signal, the interaction confidence indicating a reliability that the audio signal is taken as the speech instruction for interacting with the speech interaction terminal;
determining a match condition between the recognition text and the audio signal, the match condition indicating a level that the recognition text correctly reflects information comprised in the audio signal; and
obtaining the wake-up word result and the speech instruction result based on the interaction confidence, the match condition and the wake-up word detection result of the front part.

10. The method of claim 8 , wherein the method is executed by a server;
and obtaining the audio signal comprises:
receiving the audio signal sent by the speech interaction terminal.

11. A control apparatus for speech interaction, comprising:
a non-transitory computer-readable medium including computer-executable instructions stored thereon, and an instruction execution system which is configured by the instructions to implement at least one of:
a collecting module, configured to collect an audio signal;
a detecting module, configured to detect a wake-up word in the audio signal to obtain a wake-up word result; and
an executing module, configured to play a prompt tone and/or to execute a speech instruction in the audio signal based on the wake-up word result,
the wake-up word result comprises a first confidence and a second confidence, the first confidence is configured to represent a reliability that the audio signal comprises a target wake-up word, the second confidence is configured to represent a reliability that the audio signal comprises an ordinary wake-up word, the number of syllables of the target wake-up word is no more than that of the ordinary wake-up word, and
the executing module comprises:
an instruction executing module, configured to execute the speech instruction in a case that the first confidence reaches a first confidence threshold; and
a playing module, configured to play the prompt tone in a case that the second confidence reaches the second confidence thresho

Assignees

• Beijing Baidu Netcom Sci & Tech Co Ltd

Inventors

• Gao Cong
• Zou Saisai
• Bai Jinfeng
• Jia Lei

Classifications

• G10L15/30

  Distributed recognition, e.g. in client-server systems, for mobile phones or network applications · CPC title

• G10L2015/088

  Word spotting · CPC title

• G10L15/22 Primary

  Procedures used during a speech recognition process, e.g. man-machine dialogue · CPC title

• G10L2015/225

  Feedback of the input speech · CPC title

• G06F3/167 Primary

  Audio in a user interface, e.g. using voice commands for navigating, audio feedback · CPC title