Face recognition-based video recommendation method and apparatus, and television

Patent metadata

Field	Value
Publication number	US-12524985-B2
Application number	US-202017758186-A
Country	US
Kind code	B2
Filing date	Aug 25, 2020
Priority date	Dec 30, 2019
Publication date	Jan 13, 2026
Grant date	Jan 13, 2026

Abstract

Disclosed are a face recognition-based video recommendation method and an apparatus, and a television. The method includes: acquiring a face image, the face image comprising a face of at least one user (S1); acquiring, according to a face recognition algorithm, a face feature set of the at least one user from the face image (S2); comparing the face feature set with face feature sets in an archive (S3); and if the face feature set of the at least one user matches the face feature sets in the archive, displaying a first video recommendation set, and if each face feature set does not match the face feature sets in the archive, display a second video recommendation set (S4). Using the method to recommend videos to users reduces operation steps of the users choosing to watch videos and provides convenience for the users to use.

First claim

What is claimed is:

1 . A face recognition-based video recommendation method, the method comprising:
acquiring a face image, the face image comprising a face of at least one user; acquiring a face feature set of the at least one user from the face image according to a face recognition algorithm; comparing the face feature set with face feature sets in an archive; displaying a first video recommendation set if the face feature set of the at least one user matches the face feature sets in the archive, and displaying a second video recommendation set if each face feature set does not match the face feature sets in the archive; wherein the first video recommendation set includes a first video recommendation subset and a second video recommendation subset; displaying the first video recommendation set if the face feature set of the at least one user matches the face feature sets in the archive comprises: if among the face feature set of the at least one user, a face feature set of only one user matches the face feature sets in the archive, displaying the first video recommendation subset corresponding to the only one user; and if among the face feature set of the at least one user, face feature sets of multiple users match the face feature sets in the archive, displaying the second video recommendation subset corresponding to the face feature sets of the multiple users; wherein a process of the face recognition algorithm comprises: using a facial detector to detect faces in an acquired face image; creating objects of a Facemark class, loading a facial landmark detector, wherein the facial landmark detector is trained on thousands of face images with landmark labels and obtained; running the facial detector on the acquired face image, wherein an output of the facial detector is a vector containing one or more rectangles, and the face image comprises one or more faces; capturing a face ROI of an original image according to a face rectangular box, and then using the facial landmark detector to detect the face ROI, for each face, multiple landmarks are obtained and stored; and according to the landmarks, drawing the landmarks on the face image and display the landmarks; wherein if the face feature set of the at least one user is a1, pre-stored face feature sets of multiple users comprise b1 and b2, if a degree of matching between a1 and b1 is greater than a threshold value, the first video recommendation set corresponding to b1 is played; if the degree of matching between a1 and b1 is less than the threshold value, and the degree of matching between a1 and b2 is also less than the threshold value, the second video recommendation set is displayed; wherein after displaying the second video recommendation set, an account is created for a user whose face feature set does not match any in the archive, and the user's face feature set is associated with the account and stored in the archive; wherein before acquiring the face image, the method further comprises determining a user favorite video set based on a view history in a user account using a big data algorithm.

2 . The face recognition-based video recommendation method as claimed in claim 1 , wherein before acquiring the face image, the method further comprises: establishing the archive in advance.

3 . The face recognition-based video recommendation method as claimed in claim 2 , wherein establishing the archive in advance further comprises:
acquiring face images of users; obtaining face feature sets of the users from the face images of the users according to the face recognition algorithm; and storing the face feature sets of the users in the archive, wherein the face feature sets of the users correspond to accounts of the users.

4 . The face recognition-based video recommendation method as claimed in claim 1 , wherein acquiring the face feature set of the user from the face image according to the face recognition algorithm further comprises:
processing the face image to obtain a processed face image; performing face feature recognition on the processed face image to obtain face features of the user; and filtering the face features of the user to obtain the face feature set of the user, wherein the face feature set of the user at least includes facial features and facial contour features of the user.

5 . The face recognition-based video recommendation method as claimed in claim 4 , wherein processing the face image to obtain the processed face image comprises:
converting an acquired face image from an analog signal to a digital signal to obtain a first image; performing a binarization process on the first image to obtain a second image; performing a smoothing process on the second image to obtain a third image; and transforming the third image to obtain the processed face image, wherein transformation is used to correct systematic errors in the third image.

6 . The face recognition-based video recommendation method as claimed in claim 1 , wherein displaying the first video recommendation set if the face feature set of the at least one user matches the face feature sets in the archive comprises:
displaying the first video recommendation set if a degree of matching between the face feature set of the at least one user and the face feature sets in the archive is greater than or equal to a threshold value.

7 . The face recognition-based video recommendation method as claimed in claim 1 , wherein displaying the second video recommendation set if each face feature set does not match the face feature sets in the archive comprises:
displaying the second video recommendation set if a degree of matching between the face feature set of the at least one user and the face feature sets in the archive is less than a threshold value.

8 . The face recognition-based video recommendation method as claimed in claim 1 , wherein the first video recommendation subset corresponding to the only one user is a user favorite video set corresponding to the only one user.

9 . The face recognition-based video recommendation method as claimed in claim 1 , wherein the second video recommendation subset corresponding to the face feature sets of the multiple users is an intersection of multiple user favorite video sets corresponding to the multiple users.

10 . The face recognition-based video recommendation method as claimed in claim 1 , wherein the second video recommendation set is a public favorite video set.

11 . The face recognition-based video recommendation method as claimed in claim 1 , wherein after displaying the second video recommendation set if each face feature set does not match the face feature sets in the archive, the method further comprises:
creating an account for the at least one user; and storing the face feature set of the at least one user to the archive, the face feature set of the at least one user associated with the account of the at least one user.

12 . A television, the television comprising: a memory and a processor, the memory storing a computer program, wherein the processor implements the method as claimed in claim 1 when executing the computer program.

13 . A non-transitory computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the method as claimed in claim 1 are implemented.

Assignees

• Shenzhen Tcl Digital Tech Ltd

Inventors

• Li Rongkun

Classifications

• H04N21/466

  Learning process for intelligent management, e.g. learning user preferences for recommending movies {(services using the results of monitoring in broadcast systems H04H60/61)} · CPC title

• H04N21/251

  Learning process for intelligent management, e.g. learning user preferences for recommending movies (details of learning user preferences for the retrieval of video data in a video database G06F16/739; computer systems using learning methods G06N3/08) · CPC title

• G06V40/161

  Detection; Localisation; Normalisation · CPC title

• G06V40/172

  Classification, e.g. identification · CPC title

• G06V40/168

  Feature extraction; Face representation · CPC title