Unlock instant, AI-driven research and patent intelligence for your innovation.

Human motion prediction method based on adversarial training attention mechanism

A technology of human action and attention, applied in the field of human-computer interaction, can solve the problems of computing resource consumption, discontinuity of the first frame, etc., and achieve the effect of stable training process

Pending Publication Date: 2022-04-22
DALIAN UNIV OF TECH
View PDF0 Cites 1 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0010] The purpose of the present invention is to solve the problem of discontinuity of the first frame when predicting human actions and the computational resource consumption and discontinuity of the first frame of the attention mechanism in the Transformer model when predicting human actions

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Human motion prediction method based on adversarial training attention mechanism
  • Human motion prediction method based on adversarial training attention mechanism
  • Human motion prediction method based on adversarial training attention mechanism

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0042] The present invention will be described in further detail below in conjunction with specific embodiments. The specific embodiments described here are only used to explain the present invention, and are not intended to limit the present invention.

[0043] This embodiment discloses a human action prediction method based on the attention mechanism of confrontation training, and its detailed network structure schematic diagram is as follows figure 1 shown. Specific steps are as follows:

[0044] (1) Human body motion joint point data processing

[0045] In this embodiment, the Human3.6m data set is used, and the secondary data set contains 15 actions. To read all the data from the data set, such as "running", you need to traverse all the files and read the data by file. Divide the human body into 32 joint points, and use the sequence X={x 1 ,x 2 ,...,x t}∈R T×NF where T represents the time range; N is the number of human joints; F represents the dimension of the rep...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention belongs to the technical field of human-computer interaction, and relates to prediction of human body actions in human-computer interaction, in particular to a human body action prediction method based on an adversarial training attention mechanism. According to the method, on the basis of an original Transform model, transformation optimization is carried out on a Transform internal attention computing mechanism, and a deformable Transform model is designed and used for extracting time features and space features of human body movement, so that the mutual dependency relationship among all joint points in a long-time range is captured, and the human body movement in the long-time range is efficiently predicted. Secondly, an adversarial training mechanism is introduced to train a proposed network model, the process of generating motion prediction is used as a generator, and a continuity discriminator and an authenticity discriminator are introduced to verify the time smoothness and continuity of a generated sequence, so that the problem of first frame discontinuity is relieved.

Description

technical field [0001] The invention belongs to the technical field of human-computer interaction, and relates to the prediction of human action in human-computer interaction, in particular to a method for predicting human action based on the attention mechanism of confrontation training. Background technique [0002] In recent years, with the rapid development of artificial intelligence technology in the computer field, the research of human-computer interaction has attracted more and more researchers' attention. Humans have the ability to predict the surrounding dynamic environment in real time. How to make robots imitate human's prediction ability has become one of the research hotspots in the field of human-computer interaction. In a natural and efficient human-computer interaction process, only when the robot perceives the surrounding environment in time can it complete the interaction safely, thereby planning and executing follow-up tasks. Therefore, accurate predicti...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
IPC IPC(8): G06N3/04G06N3/08
CPCG06N3/04G06N3/08G06N3/047G06N3/044
Inventor 张强范宣哲于华候亚庆周东生
Owner DALIAN UNIV OF TECH