Individuals with severe physical disabilities and minimal motor behaviour may benefit from access technologies that harness the volitional activity of muscles. In this thesis, we investigate the use of muscle activity, indicated by the mechanomyogram (MMG), as a binary and multifunction control signal for access devices.
A challenge in the design of a binary MMG switch is to reliably recognize the timing of voluntary muscle contractions, and to subsequently translate the MMG signal into a switch-activation signal. A continuous wavelet transform (CWT) algorithm based on a simplified MMG generation model of recurring impulsive morphological patterns was proposed for automatic detection of muscle activity from MMG signals. CWT coefficients of the MMG signal were compared to scale-specific thresholds derived from the baseline signal to estimate the timings of muscle activity. The automatic detection algorithm was implemented as a binary switch controlled by a single muscle site at the forehead, shoulders or forearm. The binary control algorithm was verified on able-bodied participants, and the switch was tested on individuals with neuromuscular and neurological disabilities. The switch showed very high sensitivity and specificity when the muscle and its control were minimally affected by spasticity, involuntary movement, or involuntary muscle activity.
We further investigate the potential of improving the functionality of the MMG-controlled switching interface. We demonstrate the practical use of the vibratory artefact measured at the forehead during vocalization to control a second switch. The proposed integrated MMG-vocalization access solution may augment access alternatives for individuals with physical disabilities using a single access site. Further, we show that multi-site MMG signals exhibit distinctive patterns of forearm muscle activity, and 7+/-1 hand movements can be identified with an accuracy of 90+/-4%. This suggests that MMG may have applications in multi-function body-machine interfaces when multiple muscle sites are available. However, MMG signal features vary with sensor location. We show that sensor displacements significantly diminish classification accuracy, emphasizing the importance of consistent sensor placement between MMG classifier training and deployment for accurate control of switching interfaces.
Collectively, the findings and developments of this thesis lay the foundation for future research on wearable, MMG-driven access technologies.
| Identifer | oai:union.ndltd.org:LACETR/oai:collectionscanada.gc.ca:OTU.1807/25923 |
| Date | 14 January 2011 |
| Creators | Alves-Kotzev, Natasha |
| Contributors | Chau, Tom |
| Source Sets | Library and Archives Canada ETDs Repository / Centre d'archives des thèses électroniques de Bibliothèque et Archives Canada |
| Language | en_ca |
| Detected Language | English |
| Type | Thesis |
Page generated in 0.0027 seconds