E. S. Gopi

Digital image forgery detection using artificial neural network and independent component analysis

Digital image forgery is the process of manipulating the original photographic images like resizing, rotation, scaling, etc. To produce the photographic images as the evidence to the court, there is the need to identify whether the produced image is original or forgery image. In this paper, an attempt is made to detect forgery portions of the digital image. This is achieved by training the artificial neural network using the ICA coefficients obtained in the AR domain of the image data.

Formulating particle swarm optimization based membership linear discriminant analysis

Abstract In this paper, we propose the technique for computing the scatter matrices that are used in the linear discriminant analysis using the modified centroids. Modified (weighted) centroids are computed as the weighted mean of all the vectors belonging to the identical cluster. The membership values which are used to compute the weighted centroids are obtained using the fuzzy membership values and the particle swarm optimization. The experiments are performed using the 2-D, 3-D toy clusters and the real database. The experiment results reveal the fact that the detection rate has increased significantly when the classical centroids are replaced with the weighted centroids in LDA.

Fast computation of PCA bases of image subspace using its inner-product subspace

Computation of Principal Component Analysis (PCA) basis for Image subspace requires more memory and takes comparatively more time for computation. In this paper a simple approach to obtain the PCA basis of the Image subspace using its inner-product space is proposed. Experimental results depicting the reduction of computation time using the proposed technique is reported.

Music Indexing Using Independent Component Analysis with Pseudo-generated Sources

In this paper we present a new approach towards Singing Voice/ Music segmentation using Independent Component Analysis. If the singing voice and the background music are assumed to be two independent signals mixed to form the song, Independent Component Analysis can be used to separate them. ICA requires at least two sources in order to separate two mixed signals, whereas in this case only a single source, i.e. the recording of the song, is available. Another pseudo source is generated from the single source using Discrete Wavelet Transform and the discrimination between singing voice and music is done using a Feed Forward Back Propagation Neural Network.

Neural network based class-conditional probability density function using kernel trick for supervised classifier

The practical limitation of the Bayes classifier used in pattern recognition is computing the class-conditional probability density function (pdf) of the vectors belonging to the corresponding classes. In this paper, a neural network based approach is proposed to model the class-conditional pdf, which can be further used in supervised classifier (e.g. Bayes classifier). It is also suggested to use kernel version (using kernel trick) of the proposed approach to obtain the class-conditional pdf of the corresponding training vectors in the higher dimensional space. This is used for better class separation and hence better classification rate is achieved. The performance of the proposed technique is validated by using the synthetic data and the real data. The simulation results show that the proposed technique on synthetic data and the real data performs well (in terms of classification accuracy) when compared with the classical Fisher?s Linear Discriminant Analysis (LDA) and Gaussian based Kernel-LDA.

Statistical Signal Processing

This chapter deals with the generating model of the signal like speech signal using AR, MA and ARMA. This chapter covers how the adaptive filters are used for system model and noise removal. It also deals with estimating the spectral content of the given signal using eigen decomposition, Pisarenko Harmonic decomposition, MUSIC and ESPRIT method.

Particle Swarm Optimization Based HMM Parameter Estimation for Spectrum Sensing in Cognitive Radio System

Spectrum Estimation has emerged as the major bottleneck for the development of advanced technologies (IoT and 5G) that demand for a unperturbed continuous availability of the spectrum resources. Opportunistic dynamic access of spectrum by unlicensed users when the licensed user is not using the resources is seen as a solution to the pressing issue of spectrum scarcity. The idea proposed for spectrum estimation is to model the Cognitive Radio (CR) network as Hidden Markov Model (HMM). The spectral estimation is done once in a frame. 100 such frames with 3000 slots each is considered for performing the experiment, assuming that the PU activity is known for a fraction of \(3.33\%\) of the slots i.e., for 100 slots. The parameters of the HMM are estimated by maximizing the generating probability of the sequence using the Particle Swarm Optimization (PSO). For the typical values of the network parameters, the experiments are performed and the results are presented. A novel sum squared error minimization based “Empirical Match” algorithm is proposed for an improved latent sequence estimation.

Scatter Matrix versus the Proposed Distance Matrix on Linear Discriminant Analysis for Image Pattern Recognition

In this paper, we explore the performance of Linear Discriminant Analysis (LDA) by replacing the scatter matrix with the distance matrix for image classification. First we present the intuitive arguments for using the distance matrix in LDA. Based on the experiments on face image database, it is observed that the performance in terms of prediction accuracy is better when the distance matrix is used instead of scatter matrix in Linear Discriminant Analysis (LDA) under certain circumstances. Above all, it is observed consistently that the variation of percentage of success with the selection of training set is less when distance matrix is used when compared with the case when scatter matrix is used. The results obtained from the experiments recommend the usage of distance matrix in place of scatter matrix in LDA. The relationship between the scatter matrix and the proposed distance matrix is also deduced.