Current solutions for the control of active upper-limb prostheses are mostly based on EMG signals acquisition and processing and on electronic switches. Even though efficient for most clinical cases, these solutions can be unsatisfactorily for the control of prostheses with multiple joints. Voice-control can be a possible solution for these clinical cases. The aims of this work were therefore 1) to identify a non-redundant vocabulary for the voice-control of an active upper-limb prosthesis, by maximising the recognition performances of the voice controller VR-STAMP (Sensory Inc., Sunnyvale California), and 2) to integrate the VRSTAMP with a prosthesis controller. A non-redundant vocabulary was identified comprising 26 words. The vocabulary was tested on 16 subjects, reporting no statistically significant differences between words recognition. The median number of recognitions per word per subject was 10/10 with an interquartile distance of 1. For the development of the voicecontrolled prosthesis, a firmware for the VR-STAMP was firstly developed; then, the VRSTAMP was interfaced via serial-port with the prosthesis controller CLC2000 developed and commonly used by the INAIL Prostheses Centre.
A Voice-Controlled Prosthesis: Test Of A Vocabulary And Development Of The Prototype
MAINARDI, Elena;
2008
Abstract
Current solutions for the control of active upper-limb prostheses are mostly based on EMG signals acquisition and processing and on electronic switches. Even though efficient for most clinical cases, these solutions can be unsatisfactorily for the control of prostheses with multiple joints. Voice-control can be a possible solution for these clinical cases. The aims of this work were therefore 1) to identify a non-redundant vocabulary for the voice-control of an active upper-limb prosthesis, by maximising the recognition performances of the voice controller VR-STAMP (Sensory Inc., Sunnyvale California), and 2) to integrate the VRSTAMP with a prosthesis controller. A non-redundant vocabulary was identified comprising 26 words. The vocabulary was tested on 16 subjects, reporting no statistically significant differences between words recognition. The median number of recognitions per word per subject was 10/10 with an interquartile distance of 1. For the development of the voicecontrolled prosthesis, a firmware for the VR-STAMP was firstly developed; then, the VRSTAMP was interfaced via serial-port with the prosthesis controller CLC2000 developed and commonly used by the INAIL Prostheses Centre.I documenti in SFERA sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
