Liveness detection method and apparatus, electronic device, and storage medium

The method according to claim 11, wherein the extracting time-frequency information from the reflected audio signal comprises:

• performing signal processing on the reflected audio signal to obtain a processed audio signal, and performing time-frequency analysis on the processed audio signal to obtain the time-frequency information of the processed audio signal;
• the processed audio signal comprises at least one audio frame signal;
• the audio feature comprises at least one audio frame feature;
• the video data comprises at least one image frame;
• the motion feature comprises at least one image frame feature;
• and the calculating first attention feature information of the object according to the audio feature, and calculating second attention feature information of the object according to the motion feature comprises:
• calculating attention information of a correlation between the at least one audio frame feature in the audio feature by using a first transformer encoder of a trained detection model to obtain the first attention feature information;
• and
• calculating attention information of a correlation between the at least one image frame feature in the motion feature by using a second transformer encoder of the trained detection model to obtain the second attention feature information.

The method according to claim 11, wherein the performing consistency verification on the reflected audio signal and the video data of the object based on the first attention feature information and the second attention feature information, and determining, according to a verification result, whether the liveness detection result is correct comprises:

• calculating mutual attention information between the first attention feature information and the second attention feature information by using a transformer decoder of the trained detection model;
• and
• performing consistency verification on the mutual attention information by using a verification network of the trained detection model, to obtain a verification result;
• and
• determining, when the verification result succeeds, that the liveness detection result is correct.

A liveness detection apparatus, comprising:

• at least one memory storing a computer program, and at least one processor coupled to the memory, the at least one processor being configured to execute the computer program and perform:
• obtaining a reflected audio signal and video data of an object to be detected in response to receiving a liveness detection request;
• extracting time-frequency information from the reflected audio signal;
• extracting motion trajectory information of the object from the video data;
• extracting features respectively from the time-frequency information and the motion trajectory information to obtain an audio feature and a motion feature of the object;
• calculating, according to the audio feature, first global attention information of the object indicating a global correlation of the audio feature, and calculating, according to the motion feature, second global attention information of the object indicating a global correlation of the motion feature;
• and
• determining a liveness detection result of the object by fusing the first global attention information with the second global attention information.

A non-transitory computer-readable storage medium, storing a plurality of instructions, the instructions being loaded by at least one processor and causing the at least one processor to perform:

• obtaining a reflected audio signal and video data of an object to be detected in response to receiving a liveness detection request;
• extracting time-frequency information from the reflected audio signal;
• extracting motion trajectory information of the object from the video data;
• extracting features respectively from the time-frequency information and the motion trajectory information to obtain an audio feature and a motion feature of the object;
• calculating, according to the audio feature, first global attention information of the object indicating a global correlation of the audio feature, and calculating, according to the motion feature, second global attention information of the object indicating a global correlation of the motion feature;
• and
• determining a liveness detection result of the object by fusing the first global attention information with the second global attention information.