Ivan Vilovic

An Experience in Image Compression Using Neural Networks

There are number trials of using neural networks as signal processing tools for image compression. In this paper, a direct solution method is used for image compression using the neural networks. An experience of using multilayer perceptron for image compression is presented. The multilayer perceptron is used for transform coding of the image. The network is trained for different number of hidden neurons with direct impact to compress ratio. It is experimented with different images that have been segmented in the blocks of various sizes for compression process. Reconstructed image is compared with original image using signal-to-noise ratio and number of bits per pixel. The results show the possibility of using multilayer perceptrons for image compression

PSO and ACO algorithms applied to location optimization of the WLAN base station

The main goal of this work is to show the use of evolutionary computation techniques. The particle swarm optimization (PSO) and ant colony optimization (ACO) in indoor propagation problem. These algorithms employ different strategies and computational efforts, but also they have something in common. Therefore, it is appropriate to compare their performance with the genetic algorithm (GA). We have demonstrated their ability to optimize base station location using data from neural network model of wireless local area network (WLAN). The results show that PSO has- better properties compared to ACO algorithm. The ACO algorithm needs further work to optimize the algorithm parameters, improve analysis of pheromone data and reduce computation time. However, the ant colony based approach is utilizable for solving such problems.

Performance of the Bluetooth-based WPAN for multimedia communication

In recent years, wireless ad hoc networks have been a growing area of research. Especially a Bluetooth-base wireless personal area network (WPAN) for multimedia communication requires more attention. The WPAN is new standard under development, as a part of the IEEE802.15 standard. This paper analyzes performance Bluetooth piconet for video transmission from theoretical and simulation point of view. The presented results shows feasibility of multimedia communication in the Bluetooth WPAN.

A comparison of neural network models for indoor field strength prediction

This paper presents a comparison of the field strength prediction in indoor environments based on ray tracing, multilayer perceptron and radial basis function networks. It has been already shown for neural networks as powerful tool in RF propagation prediction. It is very important to choose proper algorithm for training a neural network, so we compared several training algorithms for the case of multilayer perceptron model. As the case used a corridor of university building in Dubrovnik, for which calculation, simulation and measurement of signal strength were obtained. The results show an improvement in field strength prediction with neural models over conventional models if training algorithm and neural network architecture are carefully chosen. The best results are obtained by the radial basis function neural network model.

Location Optimization of WLAN Access Points Based on a Neural Network Model and Evolutionary Algorithms

In this article we intend to show the use of well-known evolutionary computation techniques—Particle Swarm Optimization (PSO) and Ant Colony Optimization (ACO) - in an indoor propagation problem. Although these algorithms employ different strategies and computational efforts, they also share certain similarities. Their performance is compared with a genetic algorithm (GA), which is used as reference in this case. The ability of these algorithms to optimize access point locations using data derived from the neural network model of a particular Wireless Local Area Network (WLAN) is demonstrated. Better results are obtained by the PSO algorithm compared to the ACO algorithm. Although the ACO algorithm requires further work to optimize its parameters, improve the analysis of pheromone data and reduce computation time, the ant colony-based approach is useful for solving propagation problems.

Neural Network Prediction of Signal Strength for Irregular Indoor Environments

A neural-network based approach for modelling propagation inside complex indoor environments is presented. Selection of the neural network model, initialization, and training and performance evaluation are studied in details. Furthermore, in order to determine optimal access point arrangement the neural network propagation model is merged with the particle swarm optimization method. In the case of simple indoor environments the developed propagation model is equally accurate as the deterministic methods, while in the case of complex environments the proposed method shows superior properties. Finally, the calculated results were tested in direct comparison with the measurements for both simple and complex indoor environments.