Mimicking hand motion using sEMG-based techniques for controlling a prosthesis in a natural and intuitive way
This project deals with the design and implementation of a framework capable of generating hand motor decoding models from electromyographic (EMG) signals. The intention is to use these decoding models in the control of hand prostheses. For this purpose, the development has been oriented based on th...
| Autor: | |
|---|---|
| Formato: | tesis de maestría |
| Fecha de publicación: | 2024 |
| País: | España |
| Recursos: | Universitat Politècnica de Catalunya (UPC) |
| Repositorio: | UPCommons. Portal del coneixement obert de la UPC |
| Idioma: | inglés |
| OAI Identifier: | oai:upcommons.upc.edu:2117/413034 |
| Acesso em linha: | https://hdl.handle.net/2117/413034 |
| Access Level: | acceso abierto |
| Palavra-chave: | Electromyography Electromiografia Àrees temàtiques de la UPC::Enginyeria biomèdica::Robòtica mèdica |
| Resumo: | This project deals with the design and implementation of a framework capable of generating hand motor decoding models from electromyographic (EMG) signals. The intention is to use these decoding models in the control of hand prostheses. For this purpose, the development has been oriented based on the paradigm of Learn from Demonstration (LfD), which is based on learning complex tasks through observation and imitation of human behaviour. The idea was to develop a system that a person with hand amputation could use to train his or her prosthesis to perform specific movements. Furthermore, this system adapts to the characteristics of the user, being independent of the muscles selected for control or the movements to be performed. With this, the system consists of two phases: a training phase in which the subject shows the movements to be performed by the prosthesis with their healthy arm while generating the control signals in their affected arm, and an evaluation phase in which the result of the decoding is compared with the expected movements. To do this, a stereo vision system was used in conjunction with markerless hand kinematics detection software. EMG signal acquisition equipment was used. A robotic hand has been built and programmed to serve as a prosthesis to be controlled in the system. Software based on a ROS environment has been developed for data management and communication. And two models have been trained for decoding: a multiple linear regression model and a Linear Discriminant Analysis (LDA) classifier. After the implementation of the system, an experimental protocol has been carried out in order to evaluate the performance of the system. In this protocol, a group of 6 volunteers without amputation have carried out tests that have allowed the validation of the system, as well as analysing the performance of the different parts that compose it. With this, it has been verified that the system is capable of generating the models proposed. It has also been seen that the two models used in this work are capable of identifying the user's movement intentions. In conclusion, the implemented system successfully meets the proposed objectives. And thanks to the modularity offered by ROS, it opens the door to future developments that integrate new models for decoding or use more advanced commercial prostheses into the system. |
|---|