Title:
BEHAVIOR IDENTIFICATION METHOD AND APPARATUS BASED ON DEEP NETWORK TECHNOLOGY, AND STORAGE MEDIUM
Document Type and Number:
WIPO Patent Application WO/2020/211243
Kind Code:
A1
Abstract:
The present invention relates to the technical field of image identification, particularly relates to a deep network system and behavior identification method represented on the basis of a Lie group skeleton, and specifically relates to a behavior identification method and apparatus based on deep network technology, and a storage medium, wherein the deep network system comprises a logarithm mapping layer, an excitation layer, a first complete connection layer, a time information learning layer, etc. From original deep skeleton data, extracted Lie group skeleton features are input into a CNN network designed for a Lie group, and transformed Lie-algebra features are input into a Bi-LSTM network (the time information leanring layer) before the complete connection layer; and then, prediction labels and scores from softmax layers (output layers) of the two networks are merged so as to effectively identify a behavior.
Inventors:
LI YANSHAN (CN)
GUO TIANYU (CN)
XIA RONGJIE (CN)
LIU XING (CN)
XU JIANJIE (CN)
GUO TIANYU (CN)
XIA RONGJIE (CN)
LIU XING (CN)
XU JIANJIE (CN)
Application Number:
PCT/CN2019/102983
Publication Date:
October 22, 2020
Filing Date:
August 28, 2019
Export Citation:
Assignee:
UNIV SHENZHEN (CN)
International Classes:
G06N3/04; G06K9/62
Foreign References:
| CN107229920A | 2017-10-03 | |||
| CN109614899A | 2019-04-12 | |||
| US20180260698A1 | 2018-09-13 | |||
| US20170344829A1 | 2017-11-30 |
Other References:
LI, YANSHAN ET AL.: "Skeleton-based Action Recognition with Lie Group and Deep Neural Networks", 2019 IEEE 4TH INTERNATIONAL CONFERENCE ON SIGNAL AND IMAGE PROCESSING, 21 July 2019 (2019-07-21), DOI: 20200102151318PX
HUANG, ZHIWU ET AL.: "Deep Learning on Lie Groups for Skeleton-based Action Recognition", 2017 IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION, 26 July 2017 (2017-07-26), XP080745183, DOI: 20200102151443A
HUANG, ZHIWU ET AL.: "Deep Learning on Lie Groups for Skeleton-based Action Recognition", 2017 IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION, 26 July 2017 (2017-07-26), XP080745183, DOI: 20200102151443A
Attorney, Agent or Firm:
EXCELLENT INTELLECTUAL PROPERTY LAW FIRM (CN)
Download PDF:
Previous Patent: BEHAVIOR RECOGNITION-BASED METHOD, APPARATUS AND STORAGE MEDIUM
Next Patent: SMART WATCH
Next Patent: SMART WATCH