Capturing biosignals from wearable devices is widely applied in the human-robot interaction (HRI) area. For example, surface electromyography (sEMG) signals are always adopted to track finger flexion for hand robot control. However, it is difficult to extract features from the weak sEMG signals with several noises. The existing regression model cannot be dealing with changes in real robot control scenarios. This paper proposed an sEMG based finger flexion tracking framework for robot hand control using the active learning strategy. It consists of an offline regression model and an online model updating module—the former aims to build the regression model based on the processed sEMG and finger angles. The latter is to update the model when it gets a trigger. The comparison results prove the performance of the active learning strategy in the online scenario. By comparing the overall updating times and errors, the decision tree method saves more computational time. At the same time, Gaussian regression obtains a higher accuracy.

Active learning strategy of finger flexion tracking using sEMG for robot hand control

Qi W.;Su H.;Zhang J.;Song R.;Ferrigno G.;de Momi E.;Aliverti A.
2021

Abstract

Capturing biosignals from wearable devices is widely applied in the human-robot interaction (HRI) area. For example, surface electromyography (sEMG) signals are always adopted to track finger flexion for hand robot control. However, it is difficult to extract features from the weak sEMG signals with several noises. The existing regression model cannot be dealing with changes in real robot control scenarios. This paper proposed an sEMG based finger flexion tracking framework for robot hand control using the active learning strategy. It consists of an offline regression model and an online model updating module—the former aims to build the regression model based on the processed sEMG and finger angles. The latter is to update the model when it gets a trigger. The comparison results prove the performance of the active learning strategy in the online scenario. By comparing the overall updating times and errors, the decision tree method saves more computational time. At the same time, Gaussian regression obtains a higher accuracy.
2021 6th IEEE International Conference on Advanced Robotics and Mechatronics, ICARM 2021
978-1-6654-3909-1
Decision trees; Human robot interaction; Learning systems; Palmprint recognition
File in questo prodotto:
File Dimensione Formato  
Active_Learning_Strategy_of_Finger_Flexion_Tracking_using_sEMG_for_Robot_Hand_Control.pdf

Accesso riservato

Descrizione: ORIGINAL RESEARCH article
: Publisher’s version
Dimensione 6.26 MB
Formato Adobe PDF
6.26 MB Adobe PDF   Visualizza/Apri

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11311/1203699
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 0
  • ???jsp.display-item.citation.isi??? 0
social impact