Volume holographic associative memory in a photorefractive crystal provides
an inherent mechanism to develop a multi-channel correlation identificatio
n system with high parallelism. Wavelet transform is introduced to improve
discrimination of the system. We first investigate parameters of the system
for parallelism enhancement, and then study multiplexing of the system on
input objects and wavelet filters. A general volume holographic wavelet cor
relation processor has a single input-object channel and a single wavelet-f
iltering channel. In other words, it can only process one input object with
one wavelet filter at a same time. Based on the fact that a volume hologra
phic correlator is not a shift-invariant system, multiplexing of input obje
cts is proposed to improve parallelism of the processor. As a result, sever
al input objects can be recognized simultaneously. Multiplexing of wavelet
filters with different wavelet parameters is also achieved by a Dammann gra
ting. Wavelet correlation outputs with different filters are synthesized to
improve recognition accuracy of the processor. Corresponding experimental
results in human face recognition are given. The combination of the input o
bject multiplexing and the wavelet filter multiplexing is also described. (
C) 2000 Elsevier Science B.V. All rights reserved.