A New CMOS Current-Mode Classifier Circuit for Statistics Applications

Abstract Very useful in many fields of applied electronics, the classifier find a lot of immediate applications in pattern recognition, artificial intelligence and statistics domains. A new current-mode CMOS classifier based on an Euclidean distance circuit will be further presented. The system is designed for a low-power low-voltage operation that could be achieved by avoiding to use any cascode stage. As a result of the current-mode functioning, the dynamic range of the proposed circuit could be strongly increased, while the operation in saturation region of all the MOS transistors allows obtaining a very good signal-to-noise ratio. In order to simplify the increasing of the number of input vectors, the circuit is characterized by a high modularity. All the resistors will be replaced by active MOS transistors for reducing the circuit area consumption. The Euclidean distance circuit error will be strongly reduced by designing the circuit in such a way that the output current to be independent on technological parameters. The SPICE simulations confirm the theoretical estimated results, showing a very good accuracy of the proposed circuit.