The motion pattern executed by the muscle is recognized by training certain classifiers with the features extracted to identify the executed motion. Many external factors, such as fatigue, electrode position etc can affect the EMG signals, leading to large changes in the value of a particular feature. The classifier is highly useful in identifying that unique feature throughout the variation in pattern due to external influences.
Support Vector Machines
Support vector machines is a classifier in which support vectors are extracted from the various classes of the training data and a model is thus created. The derived support vectors are then used as the determination factor for classification any data. The support vectors are the vectors in 3 dimensional space that lie closest to separating hyperplane, that separates a class from the other classes.
Consider the vector space as shown:
After applying an SVM classifier with a suitable kernel of the data; the support vector machines and hyperplanes are obtained as shown :
SVM classifiers are implemented at the software level with the help of toolboxes and libraries, the most common of which are :
As well as complete machine learning toolboxes that include SVMs:
For theoretical understanding :
Other commonly used classifiers include various artificial neural networks, Hidden Markov modeling ,k Nearest Neighbors etc.