Finger Movement Recognition for Virtual Keyboard Based on sEMG

Disabled forelimb has been a main obstacle for people to communicate with computer by keyboard. To achieve the interface with computer, most of them have to use voice commands to interact with it. But this kind of interface has many inconveniences, such as the inability of using it in both noisy and...

Full description

Saved in:
Bibliographic Details
Published in2021 5th International Conference on Imaging, Signal Processing and Communications (ICISPC) pp. 45 - 51
Main Authors Li, Longfei, An, Xuanyu, Geng, Danlei, Qin, Shiyi, Shen, Sheng
Format Conference Proceeding
LanguageEnglish
Published IEEE 01.07.2021
Subjects
Keyboards
Machine learning
Muscles
pattern recognition
Predictive models
sEMG
Signal processing
Support vector machines
SVM
Transforms
virtual keyboard
Online AccessGet full text

Cover

More Information
Summary:Disabled forelimb has been a main obstacle for people to communicate with computer by keyboard. To achieve the interface with computer, most of them have to use voice commands to interact with it. But this kind of interface has many inconveniences, such as the inability of using it in both noisy and quiet places as well as inadequate provision of commands compared to keyboard. Surface electromyography (sEMG)is a kind of signal that is created when muscles start to contract or the brain produces nerve impulses for muscle contraction. Also the existence of sEMG signals does not depend on the integrity of hands. This project aims to explore the possibility of exploiting sEMG signals to realize that the disabled can interact with the computer through the keyboard. We designed a series of hand movements (sEMG signals) and made them correspond one-to-one with specific strings. By detecting and recognizing the sEMG signals of muscles, the disabled can use the virtual keyboard to interact with the computer normally. In addition, we also designed a signal processing method based on multi-class support vector machine (SVM) with error-correcting output code (ECOC) and label-threshold pre-process. It helps patients in need create their own sEMG signal recognition model. Traditional signal processing method with floating window can accurately identify each action, but it will output the classification results of each action several times, which will cause great inconvenience to the use of the virtual keyboard. This new signal processing method with ECOC and pre-process greatly avoids this problem.
DOI:10.1109/ICISPC53419.2021.00016