Title:
ACTIONAL-STRUCTURAL SELF-ATTENTION GRAPH CONVOLUTIONAL NETWORK FOR ACTION RECOGNITION
Document Type and Number:
WIPO Patent Application WO/2022/088176
Kind Code:
A1
Abstract:
Methods, devices, and non-transitory computer readable storage medium are provided for recognizing a human action using a graph convolutional network (GCN). The method includes obtaining, by a device, a plurality of joint poses (410). The device includes a memory storing instructions and a processor in communication with the memory. The method also includes normalizing, by the device, the plurality of joint poses to obtained a plurality of normalized joint poses (420); extracting, by the device, a plurality of rough features using a modified spatial-temporal GCN (ST-GCN) from the plurality of normalized joint poses (430); reducing, by the device, a feature dimension of the plurality of rough features to obtain a plurality of dimension-shrunk features (440); refining, by the device, the plurality of dimension-shrunk features based on a self-attention model to obtain a plurality of refined features (450); and recognizing, by the device, a human action based on the plurality of refined features (460).
Inventors:
LI HAILIANG (CN)
LIU YANG (CN)
LI MAN TIK (CN)
LEI ZHI BIN (CN)
LIU YANG (CN)
LI MAN TIK (CN)
LEI ZHI BIN (CN)
Application Number:
PCT/CN2020/125878
Publication Date:
May 05, 2022
Filing Date:
November 02, 2020
Export Citation:
Assignee:
HONG KONG APPLIED SCIENCE & TECH RESEARCH INST CO LTD (CN)
International Classes:
G06K9/00
Domestic Patent References:
WO2007102537A1 | 2007-09-13 |
Foreign References:
CN110796110A | 2020-02-14 | |||
CN110738192A | 2020-01-31 | |||
CN111325099A | 2020-06-23 | |||
CN110390305A | 2019-10-29 | |||
CN104021573A | 2014-09-03 | |||
CN105069413A | 2015-11-18 | |||
CN111709268A | 2020-09-25 |
Attorney, Agent or Firm:
CHINA TRUER IP (CN)
Download PDF: