Files and links (1)
PDFOpen Access (License Unspecified), Open Access
Dissertation
Robust intent recognition for the control of myoelectric lower limb prostheses
Doctor of Philosophy (Ph.D.), Drexel University
1989
DOI:
https://doi.org/10.17918/00000979
Abstract
A study of the stochastic nature of the processed electromyogram (EMG) and performance of Gaussian Bayesian spatial pattern recognition systems, for the control of an above-knee (A/K) prosthesis, was conducted. The primary focus of this study was to identify, derive, and implement myoelectric signal processing methods that are capable of recognizing the functional intent of an individual over as large a range of force and span of time as possible. Data qualification procedures were applied to raw, processed, and patterns of processed EMG, coefficients of the linear discriminant functions of spatial pattern recognition systems, and both estimated and observed joint moments. The means and variances of all variables were found to be stationary on the second scale and nonstationary on the minute and hour scales. The one exception was the observed joint moment which was nonstationary in the variance only. A long term model of observed pattern vectors was developed to account for the nonstationary behavior. Patterns of processed EMG for isometric-isotonic and isometric-anisotonic, linearly increasing force muscle states were collected. Thus it was possible to test and evaluate a candidate "myoprocessor" with both discrete and continuously varying levels of knee moment data. Processed EMG that spanned an hour was ensemble averaged to form a Gaussian Bayesian reference model which minimized the average expected loss in the wide sense. Signal to noise ratios and percent correct classifications were obtained and proved superior to shorter term models. Improved performance was also obtained by developing a classifier normalization procedure that differed from the classical approach. Minute and hour scale stationarity was observed in parameters that were nonstationary in short term models. Recognition of an individual's functional intent has also been accomplished robustly for unplanned, nonlinear, neuromuscular events. In addition, Reference Models avoided the "catastrophic" condition that represents an uncontrollable state to the amputee to which short term models are prone. The feasibility of applying this research to real time estimation of magnitude and direction of intended limb function for the control of myoelectric lower limb prostheses has been assessed with special attention to accuracy, reliability, and computational efficiency.
Metrics
3 File views/ downloads
13 Record Views
Details
- Title
- Robust intent recognition for the control of myoelectric lower limb prostheses
- Creators
- Howard J. Hillstrom
- Contributors
- Gordon D. Moskowitz (Advisor) - Drexel University, [Retired Faculty]
- Awarding Institution
- Drexel University
- Degree Awarded
- Doctor of Philosophy (Ph.D.)
- Publisher
- Drexel University; Philadelphia, Pennsylvania
- Number of pages
- x, 125 pages
- Resource Type
- Dissertation
- Language
- English
- Academic Unit
- Biomedical Engineering and Science Institute (1970-1997); Drexel University
- Other Identifier
- 991014970192004721