Sample selection of multi-trial data for data-driven haptic texture modeling

In data-driven haptic texture rendering, the rendering quality is highly dependent on the quality of the input-output model training. The data in input model should be sufficient both in terms of quantity and coverage of the input space. Furthermore, the ever increasing input dimensions, to attain more realistic rendering makes the task of model building even more difficult. In order to address these problems, this paper proposes a novel sample selection algorithm. Our algorithm provides an efficient method of combining modeling data across multiple independent trials, whereby the significant model points are selected from each independent trial while the outliers are being eliminated. This study also provides a generic haptic model which equips other haptic modeling algorithms to benefit from the sample selection algorithm. The algorithm was evaluated using two isotropic and two non isotropic haptic texture datasets. The results showed that the algorithm provides upward of a two fold compression rate for model points, while at the same time the rendering quality remains unaffected.