Joint point motion prediction method and device, storage medium and equipment

A technology of motion prediction and joint points, which is applied in the field of image processing, can solve the problems of prediction error accumulation and limit the accuracy of human motion prediction, and achieve the effect of low prediction error and obvious advantages

CN113807255APending Publication Date: 2021-12-17CHINA TELECOM DIGITAL INTELLIGENCE TECH CO LTD
2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Publication Date
2021-12-17

Smart Images

  • Figure 1
    Figure 1
  • Figure 2
    Figure 2
  • Figure 3
    Figure 3
Patent Text Reader

Abstract

The embodiment of the invention provides a joint point prediction method, a storage medium and an electronic device, and the method comprises the steps: obtaining a to-be-detected joint point sequence, carrying out the initialization and DCT conversion, carrying out the node aggregation of different scales on the joint point sequence of the original scale after the DCT conversion through an encoder, performing feature extraction and cross-scale feature fusion on the joint points of each scale after aggregation to obtain joint point features of different scales, sequencing the joint point features of different scales according to the sizes of the scales, inputting the joint point features into a decoder for decoding, adding the output of the decoder to the joint point sequence of the original scale after DCT (Discrete Cosine Transform), and carrying out IDCT on the addition result to obtain a joint point prediction sequence. According to the method, the problems of error accumulation and insufficient long-term dependence capture caused by the adoption of RNNs are solved, and meanwhile, the collaborative motion relation between all joint points of the human body can be well modeled.
Need to check novelty before this filing date? Find Prior Art

Description

technical field

[0001] The present invention relates to the technical field of image processing, and in particular to a joint point motion prediction method, device, storage medium and equipment. Background technique

[0002] In recent years, due to the rapid development of autonomous driving, robotics and other fields, human motion prediction is getting more and more attention. The goal of human motion prediction is to use the historical 3D joint point motion sequence of the human body to predict the future 3D joint point motion sequence of the human body.

[0003] Traditional human motion prediction methods include Markov models and Gaussian mixture models, etc., but these methods can only describe simple periodic motions such as walking and swinging golf clubs. After entering the era of deep learning, the prediction of more complex human motion is usually solved by recurrent neural networks (RNNs) and feedforward neural networks (FNNs).

[0004] However, the shortcoming...

Examples

Embodiment Construction

[0055] In order to make the object, technical solution and advantages of the present invention clearer, the present invention will be further described in detail below in conjunction with the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments . Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts belong to the protection scope of the present invention.

[0056]In order to meet the requirement of high accuracy of joint point motion prediction for complex movements associated with limb coordinated motion, the embodiment of the present invention can adopt the time series data processing strategy of DCT to convert the time domain data to the frequency domain, and at the same time, the original scale after DCT transformation The joint point sequence of the joint point sequence is used for featur...