人工触覚によるパターン認識

抄録

This paper proposes an artificial-tactile pattern recognition which is composed of the recognition by touching the object surface with the artificial-tactile sense and the recognition by grasping the object with the artificial-hand.
The hint of this proposition was found in the function of the tactile sense of a human hand.
The fundermental principle of the artificial-tactile pattern recognition is to process a stress distribution that the unknown object produces in the artificial-tactile sense elements.
In the proposed method, the 3-dimensional stress distribution is partitioned into the 2-dimensional peripheral pattern and the threshold decrement by an analogy with the threshold phenomena in the living body. The object surface is recognized as a sequence of the peripheral processing at each threshold decrement.
A simple experiment classifying cylinders and square pillars was performed by the artificialhand with on-off switches instead of the presure sense elements. As the result, a high reliability of recognition is obtained.