Surface electromyography signal action recognition method based on transfer learning and support vector machine
By combining transfer learning and support vector machines with VGG16 and ResNet50 models for feature extraction, the problems of poor accuracy and insufficient generalization ability of surface electromyography signal action recognition models in upper limb action recognition are solved, and action recognition with high accuracy and strong generalization ability is achieved.
CN116898459BActive Publication Date: 2026-07-24HOHAI UNIV CHANGZHOU
View PDF 1 Cites 0 Cited by
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- HOHAI UNIV CHANGZHOU
- Filing Date
- 2023-07-06
- Publication Date
- 2026-07-24
AI Technical Summary
Technical Problem
Existing surface electromyography (EMG) signal motion recognition models have poor accuracy and generalization ability in upper limb motion recognition, which affects the practicality of the device.
Method used
A method based on transfer learning and support vector machine was adopted. The surface electromyography signal was converted into a spectrogram through short-time Fourier transform. The VGG16 and ResNet50 models were combined for feature extraction, and the concatenated features were input into the support vector machine for classification.
Benefits of technology
This improved the accuracy of upper limb motion recognition and the generalization ability of the model, ensuring the practicality of the device.
✦ Generated by Eureka AI based on patent content.
Smart Images
Figure CN116898459B_ABST
Abstract
The application discloses a surface electromyogram signal action recognition method based on transfer learning and a support vector machine in the technical field of upper limb action recognition, and aims to solve the problems that the accuracy is poor when the prior art recognizes upper limb actions, and the generalization ability of the model is poor. It comprises the following steps: collecting surface electromyogram signals corresponding to actions, and pre-processing the collected surface electromyogram signals; converting the pre-processed surface electromyogram signals into corresponding frequency spectrum graphs through short-time Fourier transform, and splicing the corresponding frequency spectrum graphs along the vertical direction to obtain a data set; inputting the obtained data set into a pre-trained classification model based on transfer learning and a support vector machine, outputting a classification result corresponding to the surface electromyogram signals through the classification model; the application is suitable for action recognition, and the method has high accuracy when recognizing upper limb actions, strong model generalization ability, and high practicability.
Need to check novelty before this filing date? Find Prior Art