Transport-based pattern recognition versus deep neural networks in underwater OAM communications.

Comparisons between machine learning and optimal transport-based approaches in classifying images are made in underwater orbital angular momentum (OAM) communications. A model is derived that justifies optimal transport for use in attenuated water environments. OAM pattern demultiplexing is performed using optimal transport and deep neural networks and compared to each other. Additionally, some of the complications introduced by signal attenuation are highlighted. The Radon cumulative distribution transform (R-CDT) is applied to OAM patterns to transform them to a linear subspace. The original OAM images and the R-CDT transformed patterns are used in several classification algorithms, and results are compared. The selected classification algorithms are the nearest subspace algorithm, a shallow convolutional neural network (CNN), and a deep neural network. It is shown that the R-CDT transformed images are more accurate than the original OAM images in pattern classification. Also, the nearest subspace algorithm performs better than the selected CNNs in OAM pattern classification in underwater environments.

Machine learning-aided classification of beams carrying orbital angular momentum propagated in highly turbid water: publisher's note.

This publisher's note corrects the name of an author of J. Opt. Soc. Am. A37, 1662 (2020)JOAOD60740-323210.1364/JOSAA.401153.

Machine learning-aided classification of beams carrying orbital angular momentum propagated in highly turbid water.

A set of laser beams carrying orbital angular momentum is designed with the objective of establishing an effective underwater communication link. Messages are constructed using unique Laguerre-Gauss beams, which can be combined to represent four bits of information. We report on the experimental results where the beams are transmitted through highly turbid water, reaching approximately 12 attenuation lengths. We measured the signal-to-noise ratio in each test scenario to provide characterization of the underwater environment. A convolutional neural network was developed to decode the received images with the objective of successfully classifying messages quickly. We demonstrate near-perfect classification in all scenarios, provided the training set includes some images taken under the same underwater conditions.

Singular value decomposition approach to coherent averaging in digital holography.

We present a new approach to coherent averaging in digital holography using singular value decomposition (SVD). Digital holography enables the extraction of phase information from intensity measurements. For this reason, SVD can be used to statistically determine the orthogonal vectors that align the complex-valued measurements from multiple frames and group common modes accounting for constant phase shift terms. The SVD approach enables the separation of multiple signals, which can be applied to remove undesired artifacts such as scatter in retrieved images. The advantages of the SVD approach are demonstrated here in experiments through fog-degraded holograms with spatially incoherent and coherent scatter.

Designing laser beams carrying OAM for a high-performance underwater communication system.

We present a design methodology for creating a distinct laser beam set suitable for detection by using only the recorded intensity pattern. We consider four coherent Laguerre-Gaussian beams carrying orbital angular momentum (OAM) to form the basis for optical communication. The complex electric fields of the beams are superimposed to create 16 dissimilar intensity patterns. The presented beam set design method considers the beam generation hardware limitations and aims to minimize the correlation among the messages and maximize their intensity differences. After propagating the 16 messages through a water channel, we measured high correlation, intensity similarity, and R-squared values for the identical messages and low values for the different ones. Distinct clustering between the measurements for the matching messages and the rest allows us to set a threshold in the gap among the groupings and successfully classify the received images.

Efficient multibeam large-angle nonmechanical laser beam steering from computer-generated holograms rendered on a liquid crystal spatial light modulator.

Multibeam large-angle beam steering is demonstrated in the visible spectral region by imprinting computer-generated holographic Fresnel zone plates on a liquid crystal spatial light modulator (SLM) configured as the first element of a telescope. The position and intensity of each beam are controlled independently. The laser beam is steered over a ±37° field of regard, with the power in the beam at 37° being greater than 50% of the on-axis power. The power delivered on axis for a single beam was 48% of the power incident on the SLM. The beam profile remained Gaussian over the full steering range, and the on-axis beam divergence is 2.1 mrad.