Unsupervised Image Clustering Using Improved Gravitational Search Algorithm
Alireza
Sardar
Department of Electrical Engineering, Faculty of Engineering, Birjand University
author
Seyed-Hamid
Zahiri
Department of Electrical and Computer Engineering, Faculty of Engineering, University of Birjand, Shoukat-Abad, Birjand, Iran.
author
text
article
2012
per
Gravitational Search Algorithm (GSA) is a novel searching and optimization algorithm which has been reported recently. GSA was inspired by the gravitational forces between the mechanical objects. The movements of the searching objects in this method are based on the estimated accelerations and velocities of them. In this paper utilizing of GSA is investigated for unsupervised image clustering. At first an improvement for the conventional GSA is presented and then an appropriate fitness function is defined for unsupervised image clustering. Extensive experimental results on various data and images demonstrate the high performance of the proposed method in comparison to other algorithms.
Journal of Soft Computing and Information Technology
Babol Noshirvani University of Technology
2383-1006
1
v.
2
no.
2012
3
18
https://jscit.nit.ac.ir/article_67322_57867e3691259144e2cb55e0cc8cebd0.pdf
Modulation Identification of Satellite Signals Using an Intelligent System
Ataollah
Ebrahimzadeh
Department of Electrical Engineering & Computer, Babol University of Technology, Babol, Iran
author
Abdollah
Doosti Aref
Department of Electrical Engineering & Computer, Babol University of Technology, Babol, Iran
author
text
article
2012
per
Automatic signal type identifier plays an important role for modulation identification of satellite signals. Most of the proposed methods didn’t have good performance in low level of signal to noise ratios (SNRs). Also they can't identify more digital modulations. This study investigates the design of an accurate system for identification of digital modulations. First, it is introduced an efficient system that includes two main modules: the feature extraction module and the classifier module. First module extracts a suitable combination of the higher order moments up to eighth, higher order cumulants up to eighth. In the classifier module, an efficient supervised classifier, i.e. radial basis function neural network is proposed. The results show this system has good performance and recognize a lot of digital modulations. However the performance of system degrades at very low SNRs. Also selection of the parameters of the classifier and feature selection is made by trial and error method. The tradeoff between them is a difficult problem. Then at the second fold we have proposed a hybrid intelligent system which an optimization module, i.e. bees algorithm (BA) is considered in the previous system.This module optimizes the classifier design by searching for the best value of the parameters and the best subset of features that feed the classifier. Simulation results show that the proposed hybrid intelligent system has very high identification accuracy even at very low SNRs. This high efficiency is achieved with little features, which have been selected using BA.
Journal of Soft Computing and Information Technology
Babol Noshirvani University of Technology
2383-1006
1
v.
2
no.
2012
63
71
https://jscit.nit.ac.ir/article_77127_05cd7cff20dae469d31eab12b7ab4024.pdf