A method for processing, recognizing and analyzing A-mode ultrasound signals of dynamic forearm gestures

By converting one-dimensional A-mode ultrasound signals into two-dimensional grayscale images and using the Vision Transformer attention mechanism network model, the problem of difficulty in analyzing the relationship between muscle movement state and gesture in A-mode ultrasound gesture recognition is solved, achieving high-precision and interpretable dynamic gesture recognition.

CN122090153APending Publication Date: 2026-05-26FUZHOU UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
FUZHOU UNIV
Filing Date
2026-02-14
Publication Date
2026-05-26

AI Technical Summary

Technical Problem

Existing technologies for gesture recognition based on A-mode ultrasound signals lack effective methods for analyzing the relationship between muscle movement states and corresponding dynamic gestures, and traditional models struggle to explain the internal data flow.

Method used

A network model based on the Vision Transformer attention mechanism is used to convert one-dimensional A-mode ultrasound signals into two-dimensional grayscale images. Spatiotemporal parameter information is generated through data preprocessing and model training, and the relationship between muscle tissue and dynamic gestures is analyzed by the attention roll-out method.

Benefits of technology

It achieves high-precision recognition and interpretable analysis of dynamic forearm gestures, improves the robustness and anti-interference ability of the model, and can accurately identify the relationship between muscle tissue changes and dynamic gestures.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122090153A_ABST
    Figure CN122090153A_ABST
Patent Text Reader

Abstract

This invention proposes a method for processing, recognizing, and analyzing A-mode ultrasound signals of dynamic forearm gestures, comprising the following steps: Step 1: Gesture data acquisition and preprocessing, sorting and organizing the A-mode ultrasound signals into a two-dimensional grayscale image according to the order of sample labels and time series; generating spatiotemporal parameter information; Step 2: Construction of a classification and recognition model, segmenting and labeling the positions of the two-dimensional grayscale image, and feeding it into the model training; Step 3: In the trained attention mechanism network model, calculating the weights of the corresponding dynamic gesture results with respect to the position of each image block using the attention rolling unfolding method, combined with the features of each image block in the two-dimensional grayscale image; obtaining a weighted heatmap of the relationship between the muscle tissue change state reflected at each position and the corresponding dynamic gesture, used to recognize the corresponding dynamic gesture; forming an interpretable and analytical scheme for gesture recognition based on A-mode ultrasound signals; This invention can provide an interpretable and analytical scheme for gesture recognition based on A-mode ultrasound signals.
Need to check novelty before this filing date? Find Prior Art