Name: JOHN JAIRO VILLAREJO MAYOR
Publication date: 03/03/2017
Advisor:
Name | Role |
---|---|
ANSELMO FRIZERA NETO | Co-advisor * |
TEODIANO FREIRE BASTOS FILHO | Advisor * |
Examining board:
Name | Role |
---|---|
ANDRE FERREIRA | Internal Examiner * |
ANSELMO FRIZERA NETO | Co advisor * |
FRANSERGIO LEITE DA CUNHA | External Examiner * |
KLAUS FABIAN COCO | External Examiner * |
TEODIANO FREIRE BASTOS FILHO | Advisor * |
Summary: Intuitive prosthesis control is one of the most important challenges to
reduce the user effort in learning how to use an artificial hand. This work presents the analysis of pattern recognition techniques for low-level myoelectric signals able to discriminate dexterous hand and fingers movements using a reduced number of electrodes in amputees. Ten amputees and ten able-bodied subjects were evaluated and the performance of the techniques was evaluated in both groups of subjects. The techniques here proposed were analyzed to classify individual finger flexion, hand movements and different grasps using four electrodes and taking into account the low level of muscle contraction in these movements. Seventeen features of myoelectric signals were also analyzed considering both traditional magnitude-based features and more recent techniques based on fractal analysis. A comparison was computed for all the techniques using different set of features, for both groups of subjects (able-bodied and amputees) with significant level of 95%. The results with a selected set of features showed average accuracy up to 92.7% of recognition for amputees using support vector machine (SVM), followed very closely by K-nearest neighbors (KNN). The results with the best combination of the analyzed techniques show that the techniques here proposed are suitable for accurately controlling dexterous prosthetic hand/fingers, providing more functionality and better acceptance for amputees.