A design method for signal processing in measurement instruments by neural networks

This paper deals with a new design method for signal processing in measurement instruments. The operations to be performed on the input signal are represented by simple mathematical and control function blocks in a flow diagram. The paper shows how each block can be replaced by means of appropriately connected neurons. The theoretical analysis and the establishment of design criteria are also given. Some applications, using programmable analog neural network chips in the instrumentation field are finally discussed.