Electronic Theses and Dissertations (Masters)

Permanent URI for this collectionhttps://hdl.handle.net/10539/37969

Browse

Now showing 1 - 3 of 3

Feasibility of region of interest selection preprocessing using a multi-photodiode fingerprint-based visible light positioning system
(University of the Witwatersrand, Johannesburg, 2024) Achari, Dipika; Cheng, Ling
This research presents a novel Multi-Photodiode Fingerprint-Based Visible Light Positioning (VLP) system aimed at improving the accuracy and reducing the computational expenses of indoor localization. The system leverages an advanced K-Nearest Neighbors (KNN) algorithm, enhanced by Signal Strength Clustering, alongside a region selection strategy based on frequency-modulated VLC encoded IDs. Through extensive simulations, the system demonstrated a notable reduction in Mean Absolute Error (MAE) to approximately 2.5 meters, with a Root Mean Square Error (RMSE) of around 3.0 meters. In addition, the system exhibited robustness across varying ambient light conditions and room sizes, maintaining an accuracy rate of 95%, even in challenging environments. Analysis revealed that error rates increased in larger rooms, with average errors ranging from 1.50 meters in smaller spaces to 3.51 meters in larger environments. This suggests that while the system is effective in smaller areas, its accuracy diminishes slightly as room size expands. However, integrating frequency domain analysis and region of interest (ROI) selection proved to be a practical approach, enhancing the overall performance of the VLP system by providing faster and more accurate indoor navigation. Future research includes exploring advanced modulation techniques integrating supplementary sensing technologies and fine-tuning the algorithm parameters to improve the system’s accuracy and reliability, especially in more complex or dynamic environments.
Model Propagation for High-Parallelism in Data Compression
(University of the Witwatersrand, Johannesburg, 2023-10) Lin, Shaw Chian; Cheng, Ling
Recent data compression research focuses on the parallelisation of existing algorithms (LZ77, BZIP2 etc.) by exploiting their inherent parallelism. Little work has been performed on parallelising highly sequential algorithms, whose slow compression speeds would benefit the most from parallelism. This dissertation presents a generalised parallelisation approach that can be potentially adopted by any compression algorithms, with model sequentiality in mind. The scheme presents a novel divide-and-conquer approach when dividing the data stream into smaller data blocks for parallelisation. The scheme, branching propagation, is implemented with prediction by partial matching (PPM), an algorithm of the statistical-modelling family known for their serial nature, which is shown to suffer from compression ratio increases when parallelised. A speedup of 5.2-7x has been achieved at 16 threads, with at most a 6.5% increase in size relative to serial performance, while the conventional approach showed up to a 7.5x speedup with an 8.0% increase. The branching propagation approach has been shown to offer better compression ratios over conventional approaches with increasing parallelism (a difference of 11% increase at 256 threads), albeit at slightly slower speeds. To quantify the speedup over ratio penalty, an alternate metric called speedup-to-ratio increase (SRI) is used. This shows that when serial dependency is maintained, branching propagation is superior in standard configurations, which offers substantial speed while minimising the compression ratio penalty relative to the speedup. However, at lower serial dependency, the conventional approach is generally preferable, with 9-16x speedup per 1% increase in compression ratio at maximal speed compared to branching propagation’s 6-13x speedup per 1%.
Modelling OAM Crosstalk with Neural Networks: Impact of Tip/tilt and Lateral Displacement
(University of the Witwatersrand, Johannesburg, 2024) Makoni, Steven Gamuchirai; Cheng, Ling
This research focuses on a critical challenge within Free Space Optical ( FSO) commu- nication systems, specifically those utilizing Mode Division Multiplexing (MDM) with Orbital Angular Momentum ( OAM ) modes of a limited transmission range. Despite these systems’ potential to significantly enhance spectral efficiency and transmission capacity, their effectiveness is hindered by the limited range caused by atmospheric turbulence-induced aberrations. Atmospheric turbulence and mis- alignments distort the optical wavefront, causing degradation in orthogonal spatial modes and resulting in power spreading into adjacent modes, known as crosstalk in MDM systems. This research presents a simple neural network model for estimating OAM crosstalk in FSO systems, specifically focusing on atmospheric turbulence-induced aberrations. Firstly, we generated datasets through simulation and experimentation for validation purposes. We then develop and evaluate the neural network model, assessing its accuracy under various turbulence aberrations. The simple neural network, trained solely on tip/tilt and displacement inputs and without retraining, accurately estimated OAM spectra using approximated inputs in turbulent condi- tions, closely matching experimentally measured spectra. Despite the presence of turbulent aberrations, the model showed a minimal decrease in the coefficient of determination, indicating its ability to generalize well to unseen measurements. Our findings indicate that a simple neural network trained solely on tilt and displacement inputs can accurately estimate OAM crosstalk amidst many turbulence aberrations for ℓ ∈ [-5, 5] as a proof of concept. This implies that simple detectors such as cameras can be used to implement or optimize digital signal processing for error detection and correction utilizing the knowledge of crosstalk, offering promising avenues for improving system efficiency and quality of service for MDM systems. In summary, this research leveraged neural networks to model OAM crosstalk induced by misalignments and turbulence. The model’s ability to estimate OAM crosstalk due to misalignments and atmospheric turbulence shows potential for use in real-time predictive systems. With further refinement, neural network models could indicate the evolution of OAM crosstalk in FSO communications that em- ploy OAM multiplexing schemes in atmospheric turbulence. The demonstrated efficacy of the neural network model positions it as a valuable tool for enhancing the robustness of FSO communications employing higher-order OAM modes.