Syntouch tactile classifier

less than 1 minute read

A developed a Qt-GUI and a set Machine Learning classifiers for the Syntouch BioTac sensor, a synthetic finger which provide pressure and temperature information at 19 different locations across the finger’s skin. Figure Data collection illustrates the data gathering step for four different classes (Corner, Edge, Surface, Air ). Once a sufficient amount of data was gathered for each class I proceeded to learn different classifiers. I found that Support Vector Machine (SVM) was the most robust classifier when compared with other classification approaches such as Gaussian Mixture Model.

Data collection:

Figure: Data collection: Gathering training data for the different classes

Figure Syntouch BioTac classifier illustratese the SVM classifier in action.

Video: Syntouch BioTac classifier Syntouch provides a 19 dimensional temporal signal. A classifier for corner, edge , surface and air was learned. The blue bars represent the probability the currently sensed data to be part of a class.