We classify surface textures using a polymer-based microelectromechanical systems (MEMS) tactile sensor array and a robust statistical approach. We demonstrate that a MEMS tactile sensor resembling a flexible sensor 'skin' built using a polyimide substrate can successfully classify textures. Texture classification is achieved by using a maximum likelihood decision rule that optimally classifies patterns in the presence of noisy signal to cope with texture variation and random noise. Using a 4 × 4 sensor array, a variety of simple textures are distinguished despite low sensitivity mechanical strain gauges serving as a transduction element. The final result analyzed using leave-one-out cross validation yields acceptable overall performance of 68% correct classification. Directions for future work to improve identification performance of the system are also presented.
ASJC Scopus subject areas
- Electronic, Optical and Magnetic Materials
- Mechanics of Materials
- Mechanical Engineering
- Electrical and Electronic Engineering