A joint torque estimation method and system based on spatio-temporal feature adaptive fusion

By constructing a joint torque estimation method that adaptively fuses spatiotemporal features, and combining temporal convolutional networks, recurrent neural networks, and robust loss functions, the accuracy and stability issues of existing methods under complex motion patterns are solved, achieving accurate capture of joint torque and improved generalization ability.

CN122140230APending Publication Date: 2026-06-05FUJIAN UNIV OF TECH
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
FUJIAN UNIV OF TECH
Filing Date
2026-01-13
Publication Date
2026-06-05

AI Technical Summary

Technical Problem

Existing joint torque estimation methods lack effective mechanisms to adaptively focus on key time segments when faced with complex motion patterns or long-term sequences. This results in limited ability to capture motion details and torque peak changes, affecting the accuracy, stability, and generalization ability of the prediction results.

Method used

A spatiotemporal feature-based adaptive fusion method is adopted. By constructing a dual-branch feature extraction structure that combines temporal convolutional networks and recurrent neural networks, and combining channel attention mechanism and temporal attention mechanism, key sensor dimension information and key time segment information that contribute significantly to torque estimation are adaptively selected and enhanced, and a robust loss function is used for training.

Benefits of technology

It significantly improves the accuracy, stability, and generalization ability of joint torque estimation, and can accurately capture motion details and torque peaks in complex motion scenarios, making it suitable for fields such as medical rehabilitation and sports science.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122140230A_ABST
    Figure CN122140230A_ABST
Patent Text Reader

Abstract

The application provides a joint torque estimation method and system based on spatiotemporal feature adaptive fusion in the technical field of biomechanical measurement and human-computer interaction, and the method comprises the following steps: step S1, obtaining a large amount of historical motion data collected by an inertial measurement unit from each target joint of different human bodies, and preprocessing each historical motion data to construct a data set; step S2, creating a joint torque estimation model based on a first feature extraction branch, a second feature extraction branch, a feature fusion module and a torque prediction module, and setting a loss function of the joint torque estimation model as a robust loss function; step S3, training the joint torque estimation model through the data set and the robust loss function; and step S4, estimating the joint torque through the trained joint torque estimation model. The application has the advantages that the accuracy, stability and generalization ability of joint torque estimation are greatly improved.
Need to check novelty before this filing date? Find Prior Art