US 12,449,524 B2
Object recognition method and apparatus based on ultrasonic echoes and storage medium
Ziyi Guo, Shenzhen (CN)
Assigned to TENCENT TECHNOLOGY (SHENZHEN) COMPANY LIMITED, Shenzhen (CN)
Filed by Tencent Technology (Shenzhen) Company Limited, Shenzhen (CN)
Filed on Nov. 22, 2022, as Appl. No. 17/992,754.
Application 17/992,754 is a continuation of application No. PCT/CN2022/071297, filed on Jan. 11, 2022.
Claims priority of application No. 202110071159.2 (CN), filed on Jan. 19, 2021.
Prior Publication US 2023/0114470 A1, Apr. 13, 2023
Int. Cl. G01S 7/539 (2006.01); G01S 15/89 (2006.01); G06V 10/75 (2022.01); G06V 10/774 (2022.01); G06V 40/16 (2022.01)
CPC G01S 7/539 (2013.01) [G01S 15/89 (2013.01); G06V 10/751 (2022.01); G06V 10/774 (2022.01); G06V 40/161 (2022.01)] 12 Claims
OG exemplary drawing
 
1. An object recognition method based on ultrasonic echoes, performed by a computer device, the method comprising:
receiving an echo signal reflected by an object and captured by a terminal in response to an activation of an application on the terminal, the echo signal corresponding to an ultrasonic signal transmitted by the terminal to the object;
extracting an ultrasonic echo feature corresponding to the echo signal, further including:
generating frequency spectrum information corresponding to the echo signal; and
processing the frequency spectrum information to obtain frequency-domain and time-domain information; and
performing feature dimension conversion on the ultrasonic echo feature using an acoustic wave conversion network to obtain a target dimension feature for characterizing the object, further including:
inputting the frequency-domain and time-domain information into the acoustic wave conversion network to obtain high-dimensional acoustic wave feature corresponding to the echo signal as the target dimension feature, wherein the acoustic wave conversion network is trained by:
transmitting single-channel ultrasonic waves to a sample object based on a preset frequency feature;
receiving reflected acoustic wave data from the sample object to determine echo training data;
collecting image training data corresponding to the sample object,
aligning the echo training data with the image training data based on execution time to generate a training sample pair;
determining a similarity loss based on the training sample pair; and
adjusting the acoustic wave conversion network according to the similarity loss;
performing image translation on the high-dimensional acoustic wave feature corresponding to the echo signal to obtain object image information corresponding to the object;
acquiring a preset image associated with a user account of the application;
comparing the object image information with the preset image to obtain comparison information; and
performing a predefined operation on the application when the comparison information indicates that the object image information matches the preset image.