Iterative generation of complex reference functions in a joint-transform correlator

Iterative learning procedures on hybrid electro-optic systems are employed to generate complex discriminant reference functions. The procedures are implemented experimentally on a joint-transform correlator by using a single inexpensive spatial light modulator. High-quality class discrimination is demonstrated even in the presence of noise. In this Letter we extend the procedure of performing the whole process of learning and pattern recognition on the same hybrid electro-optic system. 1 ' 2 We show that fully complex-valued discriminant reference functions (DRF's) can be constructed and that they are capable of high discrimination even if they are produced on inexpensive spatial light modulators (SLM's). Various architectures are suitable for the procedures to be described. The joint-transform correlator 3 (JTC) is convenient 2 because the presentation of