| Popis: |
Myoelectric signal is one of the most important bio-signals utilized in aiding and abetting physically disabled humans with the development of prosthetic devices. This work proposes and tests different algorithms for classifying finger movements with surface electromyogram (EMG) sensors. The data is collected from a state-of-the-art myoelectric sensor MyoBand, which gives an eight-channel EMG data. The eight finger motion primitives utilized are index flex, index extension, middle extension, middle flex, ring flex, ring extension, little flex, and little extension. The classifier is tested on a single male with no physical disabilities with MyoBand placed on the forearm proximal to elbow. The classifiers utilized are linear discriminant analysis (LDA) and support vector machines (SVMs) with different feature spaces. Both classifiers were implemented in MATLAB environment, and from the result analysis, the inference obtained is that LDA has the highest classification accuracy with 97.7%. However, the trade-off of the approach is that it is not tractable for real-time implementation. The SVM accentuates a better trade-off between speed and accuracy with 95.7% and is more suitable for real-time implementation. |
