Voice recognition method of artificial intelligence robot device

Patent metadata

Field	Value
Publication number	US-11508362-B2
Application number	US-202017025459-A
Country	US
Kind code	B2
Filing date	Sep 18, 2020
Priority date	Nov 19, 2019
Publication date	Nov 22, 2022
Grant date	Nov 22, 2022

Abstract

A voice recognition method of an artificial intelligence robot device is disclosed. The voice recognition method includes collecting a first voice spoken by a user and determining whether a wake-up word of the artificial intelligence robot device is recognized based on the collected first voice; if the wake-up word is not recognized, sensing a location of the user using at least one sensor and determining whether the sensed location of the user is included in a set voice collection range; if the location of the user is included in the voice collection range, learning the first voice and determining a noise state of the first voice based on the learned first voice; collecting a second voice in an opposite direction of the location of the user according to a result of the determined noise state of the first voice; and extracting a feature value of a noise based on the second voice and removing the extracted feature value of the noise from the first voice to obtain the wake-up word. The artificial intelligence robot device may be associated with an artificial intelligence module, an unmanned aerial vehicle (UAV), a robot, an augmented reality (AR) device, a virtual reality (VR) device, devices related to 5G services, and the like.

First claim

What is claimed is:

1. A voice recognition method for an artificial intelligence robot device, comprising:
collecting first audio including a voice of a user and determining whether a wake-up word of the artificial intelligence robot device is recognized based on the collected first audio;
based on the wake-up word not being recognized, sensing a location of the user and determining whether the sensed location of the user is included in a set voice collection range;
based on the location of the user being included in the set voice collection range, determining a noise state of the first audio;
collecting second audio from an opposite direction of the location of the user according to a result of the determined noise state of the first audio; and
extracting a feature value of a noise based on the second audio and removing the extracted feature value of the noise from the first audio to obtain the wake-up word from the first audio.

2. The voice recognition method of claim 1 , wherein the first audio is collected using a first microphone mounted toward a front of the artificial intelligence robot device.

3. The voice recognition method of claim 1 , wherein determining whether the wake-up word of the artificial intelligence robot device is recognized based on the collected first audio comprises:
extracting first feature values from the first audio;
inputting the first feature values to an artificial neural network (ANN) trained to recognize the wake-up word; and
determining whether the wake-up word of the artificial intelligence robot device is recognized based on an output of the artificial neural network (ANN).

4. The voice recognition method of claim 3 , wherein the first feature values are values capable of distinguishing whether the wake-up word of the artificial intelligence robot device is recognized.

5. The voice recognition method of claim 3 , wherein determining whether the location of the user is included in the set voice collection range comprises:
extracting second feature values from sensor information related to the location of the user;
inputting the second feature values to a second ANN trained to distinguish whether the location of the user is included in the set voice collection range; and
determining whether the location of the user is included in the set voice collection range based on an output of the second ANN.

6. The voice recognition method of claim 5 , wherein the second feature values are values capable of distinguishing whether the location of the user is included in the set voice collection range.

7. The voice recognition method of claim 5 , wherein determining of the noise state of the first audio comprises:
extracting third feature values of the noise from the first audio;
inputting the third feature values of the noise to a third ANN trained to distinguish noise states of inputted audio; and
determining the noise state of the first audio based on an output of the third ANN.

8. The voice recognition method of claim 7 , wherein the third feature values of the noise are values capable of distinguishing the noise state of the first audio.

9. The voice recognition method of claim 1 , wherein the second audio is collected using a second microphone mounted toward a rear of the artificial intelligence robot device.

10. The voice recognition method of claim 7 , further comprising receiving, from a network, downlink control information (DCI) that is used to schedule a transmission of the first audio or the location information,
wherein the first audio or the location information is transmitted to the network based on the DCI.

11. The voice recognition method of claim 10 , further comprising performing an initial access procedure with the network based on a synchronization signal block (SSB),
wherein the first audio or the location information is transmitted to the network via a physical uplink shared channel (PUSCH), and
wherein the SSB and a demodulation reference signal (DM-RS) of the PUSCH are quasi co-located (QCLed) for QCL type D.

12. The voice recognition method of claim 11 , further comprising:
controlling a transceiver so that the first audio or the location information is transmitted to an AI processor included in the network;
receiving information processed by the AI processor via the transceiver,
wherein the processed information is one of information for determining a recognition state of the wake-up word of the artificial intelligence robot device, information for determining whether the location of the user is included in the set voice collection range, or information for determining the noise state of the first audio.

13. The voice recognition method of claim 1 , wherein the first audio includes a wake-up word spoken by the user at an airport and a noise around the airport,
wherein the second audio includes the noise around the airport.

14. The voice recognition method of claim 1 , wherein the feature value of the noise is extracted based on an announcement of the airport, a sound of a cleaner cleaning the airport, a sound of an airplane, a sound of an airport cart, or a sound of carriers moving inside the airport.

15. The voice recognition method of claim 1 , wherein obtaining the wake-up word from the first audio comprises:
setting a noise reference range based on the extracted feature value of the noise based on the second audio; and
applying the set noise reference range to the first audio to remove the noise from the first audio to obtain the wake-up word.

16. A machine-readable non-transitory medium having stored thereon machine-executable instructions for a voice recognition method for an artificial intelligence robot device, the instructions comprising:
collecting first audio including a voice of a user and determining whether a wake-up word of the artificial intelligence robot device is recognized based on the collected first audio;
based on the wake-up word not being recognized, sensing a location of the user and determining whether the sensed location of the user is included in a set voice collection range;
based on the location of the user being included in the set voice collection range, determining a noise state of the first audio;
collecting second audio from an opposite direction of the location of the user according to a result of the determined noise state of the first audio; and
extracting a feature value of a noise based on the second audio and removing the extracted feature value of the noise from the first audio to obtain the wake-up word from the first audio.

17. The machine-readable non-transitory medium of claim 16 , wherein determining whether the wake-up word of the artificial intelligence robot device is recognized based on the collected first audio comprises:
extracting first feature values from the first audio;
inputting the first feature values to an artificial neural network (ANN); and
determining whether the wake-up word of the artificial intelligence robot device is recognized based on an output of the artificial neural network (ANN).

18. The machine-readable non-transitory medium of claim 17 , wherein determining whether the location of the user is included in the set voice collection range comprises:
extracting second feature values from information from a sensor related to the location of the user;
inputting the second feature values to a second ANN to distinguish whether the location of the user is included in the set voice collection range; and
determining whether the location of the user is included in the set voice collection range ba

Assignees

• Lg Electronics Inc

Inventors

• Lee Inho
• Lee Junmin
• Lee Keunsang

Classifications

• H04W72/23

  in the downlink direction of a wireless link, i.e. towards a terminal · CPC title

• G10L21/0208

  Noise filtering · 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

• G10L25/51

  for comparison or discrimination · CPC title