Classification of brain activity using muscle activation marker for BCI controlled hand prosthesis / Khairunnisa Johar
The use of body-powered prostheses can be tiring and lead to problems with compliance and prosthetic restoration. Apparently, brain-Computer-Interfaces (BCI) offer a means of controlling prostheses for patients that are unable to operate such devices due to physical limitations. An issue with BCIs i...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2017
|
| Subjects: | |
| Online Access: | https://ir.uitm.edu.my/id/eprint/100233/1/100233.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The use of body-powered prostheses can be tiring and lead to problems with compliance and prosthetic restoration. Apparently, brain-Computer-Interfaces (BCI) offer a means of controlling prostheses for patients that are unable to operate such devices due to physical limitations. An issue with BCIs is that they tend to either require invasive recording methods, posing a surgical risk or work by generating control signals from not task related brain activity patterns such as right versus left hand, hand versus leg or visual stimulation and therefore are not intuitive in their control. This study aims at testing the possibility of controlling the grasp and release of an upper limb prosthetic terminal device by classifying Electroencephalogram (EEG) data from isometric finger extension and flexion real movements. Data from five healthy subjects were recorded using a consumer grade non-invasive Emotiv EPOC headset. During the measurement of five subjects, they were asked to perform isometric finger extension and flexion of their right hand. To bring the EEG data into correlation with the executed movement, simultaneous electromyogram (EMG) recording was proposed as an alternative method for recording visual cued movement. Classified EMG data was used to generate markers in the EEG data and to the epoch data. The EEG data was then filtered to increase the signal to noise ratio and allow better classification while spectrally weighted common spatial patterns (spec-CSP) were used for feature extraction. Using linear discriminant analysis (LDA), a successful classification rate of up to 73.2% between isometric finger extension and flexion, 79.9% between isometric finger flexion and rest and 81.8% between isometric finger extension and rest were obtained. In this study, a novel EMG-assisted approach was developed for classifying EEG data from isometric finger extension and flexion movements. It has displayed feasibility for a more intuitive control on upper limb prosthetic terminal device using low-cost BCI without the risks associated with invasive measurement. |
|---|