Share Email Print
cover

Proceedings Paper

Fast method for dynamic thresholding in volume holographic memories
Author(s): Michael S. Porter; Pericles A. Mitkas
Format Member Price Non-Member Price
PDF $14.40 $18.00

Paper Abstract

It is essential for parallel optical memory interfaces to incorporate processing that dynamically differentiates between databit values. These thresholding points will vary as a result of system noise -- due to contrast fluctuations, variations in data page composition, reference beam misalignment, etc. To maintain reasonable data integrity it is necessary to select the threshold close to its optimal level. In this paper, a neural network (NN) approach is proposed as a fast method of determining the threshold to meet the required transfer rate. The multi-layered perceptron network can be incorporated as part of a smart photodetector array (SPA). Other methods have suggested performing the operation by means of histogram or by use of statistical information. These approaches fail in that they unnecessarily switch to a 1-D paradigm. In this serial domain, global thresholding is pointless since sequence detection could be applied. The discussed approach is a parallel solution with less overhead than multi-rail encoding. As part of this method, a small set of values are designated as threshold determination data bits; these are interleaved with the information data bits and are used as inputs to the NN. The approach has been tested using both simulated data as well as data obtained from a volume holographic memory system. Results show convergence of the training and an ability to generalize upon untrained data for binary and multi-level gray scale datapage images. Methodologies are discussed for improving the performance by a proper training set selection.

Paper Details

Date Published: 5 November 1998
PDF: 10 pages
Proc. SPIE 3468, Advanced Optical Memories and Interfaces to Computer Storage, (5 November 1998); doi: 10.1117/12.330405
Show Author Affiliations
Michael S. Porter, Colorado State Univ. (United States)
Pericles A. Mitkas, Colorado State Univ. (Greece)


Published in SPIE Proceedings Vol. 3468:
Advanced Optical Memories and Interfaces to Computer Storage
Pericles A. Mitkas; Zameer U. Hasan, Editor(s)

© SPIE. Terms of Use
Back to Top