Experiments with Self-Organizing Systems for Texture and Hardness Perception

We have experimented with different SOM-based architectures for bio-inspired self-organizing texture and hardness perception systems. To this end we have developed a microphone based texture sensor and a hardness sensor that measures the compression of the material at a constant pressure. We have implemented and successfully tested both monomodal systems for texture and hardness perception, bimodal systems that merge texture and hardness data into one representation and a system which automatically learns to associate the representations of the two submodalities with each other. The latter system employs the novel Associative Self- Organizing Map (A-SOM). All systems were trained and tested with multiple samples gained from the exploration of a set of 4 soft and 4 hard objects of different materials with varying textural properties. The monomodal texture system was good at mapping individual objects in a sensible way. This was also true for the hardness system which in addition divided the objects into categories of hard and soft objects. The bimodal system was successful in merging the two submodalities into a representation that performed at least as good as the best recognizer of individual objects, i.e. the texture system, and at the same time categorizing the objects into hard and soft. The A-SOM based system successfully found representations of the texture and hardness submodalities and also learned to associate These with each other. (Less)