PUB - Publikationen an der Universität Bielefeld

Insects use their antennae (feelers) as near-range sensors for orientation, object localization and communication. This paper presents further developments for an approach for an active tactile sensor system. This includes a hardware construction as well as a software implementationfor interpreting the sensor readings. The discussed tactile sensor is able to detect an obstacle and its location. Furthermore the material properties of the obstacles are classied by application of neural networks. The focus of this paper lies in the development of a method which allows to determine automatically the part of the input data which is actually needed to fulll the classication task. For that, non-negative matrix factorization is evaluated by quantifying the trade-off between classication accuracy and input (and network) dimension.