Options
1990
Conference Paper
Titel
Neural based multi-sensory system
Abstract
In this contribution we present an advanced concept of neural based multi-sensory systems. It realizes two important features: on the one hand the neural net is trained what the sensors are supposed to detect in a known environment. On the other hand it learns the intrinsic nonidealities of the sensors. Successfull experiments have been carried out with pressure where the network has learned the temperature dependence in addition to the sensor sensitivity. Further experiments were carried out for gas mixtures to determine the gas spectrum.