The electromyogram has numerous applications in engineering and science. One specific application is to model a system for the torque generated by the elbow joint. This application has been long studied and applied in controller designs for artificial prosthetics limbs. Previous research had shown that nonlinear and multiple channel whitened EMG signal models gave the best EMG to torque estimates compared to linear un-whitened models. This thesis describes the methodologies for predicting the torque into the future up to 1 second. Four specific types of finite impulse response models (linear and nonlinear, single channel un-whitened and multi-channel whitened) are compared based on the EMG-based predicted torque and the actual torque. The errors were measured as the difference between actual and predicted torque. It was observed that the error was mostly constant at the minimum error value between 0 and 80 ms for all four models, with the lowest error being 5.48 % maximum voluntary contraction (MVC) flexion. Further comparison was performed between different lower order models and a Butterworth second order model for predicting torque ahead in time. Such models are common in the literature. This thesis separately investigates the effect of band limiting the whitened EMG signal and using the advanced EMG processors for estimating the torque. Whitened EMG data were passed through a low pass filter with selectable cutoff frequency from 2048 Hz down to 20 Hz to limit the whitened band width. It was observed that the error was not significantly different for bandwidths down to approximately 400-600 Hz, grew gradually as the band width further decreased to 200 Hz, beyond which the error increased sharply. It can be inferred that for this particular study consisting of lower contraction levels, there is no significant power usable for whitening in the EMG signal at higher frequencies, providing an opportunity for lower sampling rate, effective noise suppression, better signal to noise ratio and implementation of low cost electrodes. This research work lead to two conference paper publications at the 2013 IEEE 39th Annual Northeast Bioengineering Conference. Two journal papers are in the writing and preparation stage which will be submitted after their completion.
| Identifer | oai:union.ndltd.org:wpi.edu/oai:digitalcommons.wpi.edu:etd-theses-1246 |
| Date | 23 April 2013 |
| Creators | Koirala, Kishor |
| Contributors | Edward A. Clancy, Advisor, , |
| Publisher | Digital WPI |
| Source Sets | Worcester Polytechnic Institute |
| Detected Language | English |
| Type | text |
| Format | application/pdf |
| Source | Masters Theses (All Theses, All Years) |
Page generated in 0.0019 seconds