Unipolar interpattern association neural networks.

A method for constructing a unipolar interconnection weight matrix (IWM) of the interpattern association (IPA) model is presented. By a search of the redundant interconnection links, a method that removes all negative links is introduced. Since the IWM is unipolar, the difficulty of implementing positive and negative IWM's can be avoided. Computer simulation and experimental results show that the unipolar IWM IPA neural network performs better than the bipolar IWM IPA model.

Multimode fiber sensing by using mean-absolute speckle-intensity variation.

We present a fiber-sensing technique that detects the mean-absolute speckle-intensity variation between the updated and the reference speckle pattern for determining the environmental perturbation factor (e.g., displacement, temperature, pressure, acoustic wave). We show that the proposed technique is highly sensitive and simple. One of the major advantages of the proposed technique is that it can perform a fast-response measurement with off-the-shelf electronic hardware. Experimental data for submicrometer displacement as well as temperature measurement are provided. To extend the dynamic range of the proposed technique, one can use an updated reference speckle pattern.

Shift-invariant interpattern association neural network.

A shift-invariant neural network that uses the translation-invariant property of the modulus Fourier spectra with the heteroassociation interpattern association memory is proposed. A binary encoding of a spectral sampling of the training set is used to preserve the main features. Computer simulations and experimental demonstrations are provided that show the shift-invariant property of the proposed optical neural network.

Implementation of a polychromatic Hamming net for color image classification.

A color exemplar-based Hamming net as applied to a color image classifier is presented. The color decomposition and composition techniques for constructing a polychromatic Hamming layer and winner-take-all (WTA) layer interconnection weight matrix (IWM) are discussed. We have simplified the WTA algorithm to reduce the number of iteration cycles in the WTA layer. To confirm the operation of the color Hamming net we provide simulation and experimental demonstrations.

Gray-level discrete associative memory.

A gray-level discrete associative-memory neural network based on object decomposition and composition is presented. By decomposing a gray-level pattern into bipolar/binary subpatterns, a gray-level discrete associative memory can be constructed from the composition of the subpattern channel results. Preprocessing for removing dc bias and normalizing the gray-level scale is performed on the input gray-level pattern. This eliminates the mismatching and saturation problems caused by bias level, which shifts the pattern gray levels throughout the pattern. Computer-simulation and optical-experimental results for a gray-level interpattern association model are shown to be consistent with the theoretical model.

Submicrometer displacement sensing using inner-product multimode fiber speckle fields.

A multimode fiber sensor using the intensity inner product of speckle fields is presented. The sensitivity and the dynamic range of the displacement sensing are quantitatively analyzed. We show that the sensitivity of displacement can be in the submicrometer range. Experimental performances show that the results are consistent with the calculated results.

Effects of thresholding in joint-transform correlation.

The joint-transform power spectrum of two identical objects can be represented as a one-dimensional sinusoidal grating modulated by a Fourier transform, and the correlation peaks can be regarded as the first-order diffraction of the grating. The peak intensity and the width are then determined by the aperture and the modulation of the grating. Based on this analysis, it is shown that dc blocking, hard clipping, or binarization of the power spectrum results in higher correlation peak intensity and a narrower peak width. Direct-current blocking is also found to be preferable if the input pattern to the correlator is corrupted by noise.