Showing papers by "Goutam Saha published in 2010"

On the use of perceptual Line Spectral Pairs Frequencies for speaker identification

Abstract: Line Spectral Pairs Frequencies (LSFs) provide an alternative representation of the linear prediction coefficients In this paper an investigation is carried out for extracting feature for speaker identification task which is based on perceptual analysis of speech signal and LSF A modified version of the standard perceptual analysis is applied to obtain better performance We have extracted the conventional LSF from the perceptually modified speech signal State-of-the art Gaussian Mixture Model (GMM) based classifier is employed to design the closed set speaker identification system The proposed method shows significant performance improvement over existing techniques in three different speech corpuses

A Robust Iterative Energy Based Voice Activity Detector

Abstract: In recent years the real time implementation of the voice biometry became a challenging field. There are several voice based technology existing today like speech recognition, speech encoding, speaker recognition, Voice over IP, voice password, hands-free telephony etc. To satisfy these requirements we need a good Voice Activity Detector (VAD) which not only detects the speech and silence portions perfectly but also the computation complexity need to be comparably lower and, the VAD should also work in different noisy environment. Our proposed VAD is time domain energy based. By iterations it calculates the threshold, still the computational complexity is found to be relatively low. In this method there is no need to predefine threshold before performing the VAD operation. The threshold parameter is updated by the SNR value which is calculated from the present speech signal. It is found to be robust and suitable for real time applications and we call it Iterative VAD.

Screening of various solvent extracts of Gymnema sylvestre R.Br. leaf for antibacterial activity.

Abstract: Increasing emergence of resistance to the currently available antibiotics has necessitated continued search for new antimicrobial compounds. The present study is aimed to confirm ethno- medicinal claim of Gymnema sylvestre leaf possessing antibacterial activity that could be a better alternative for synthetic antibacterial agents, if proved to be effective enough .For this, the antibacterial properties of Gymnema sylvestre leaf were tested against three Gram positive (Bacillus subtilis, Staphylococcus aureus and Micrococcus luteus )and five Gram negative (Escherichia coli, Vibrio cholerae, Pseudomonas aeruginosa, Shigella dysenteriae and Shigella flexneri)bacteria by using different solvents namely petroleum ether, chloroform and ethanol. The result showed that all the solvent extracts exhibited considerable activity against the tested bacteria. The antibacterial activity increased with the increasing concentration of the extract. No antibacterial activity was noted at 10 mgml -1 , 20mgml -1 .

Relevant subspace selection in Kernel feature space for speech recognition

Abstract: This paper describes an approach to select the most relevant subspace in Kernel PCA feature space applied on MFCC coefficients for speech recognition. It has been seen that the relevant information about a supervised classification problem is contained in a finite number of leading Kernel PCA components if the Kernel matches the underlying classification problem. In this paper our contribution is to foster an understanding about the appropriate Kernel selection for different phonemes in a speech database and then create an insight about the most relevant dimensions for those phonemes in that Kernel space. Using this approach we have obtained better results for speech recognition as compared to standard technique.

Determination and study of a dominant Genetic Network responsible for the growth of a fungus using the concepts of Bayesian Algorithm

Abstract: The Bayesian belief network is a powerful knowledge representation and reasoning tool under conditions of uncertainty to analyze gene expression patterns. Nowadays, this is an important tool to construct mathematical models based on probability to identify any particular dominant Genetic Network of any organism under observation. The present study deals with analysis of a set of micro array data collected at a regular interval of time throughout the growth phase of a fungus Burkholderia pseudomalli. In the first phase of the study, emphasis was given to recover a set of most dominant genes among the set of all possible expressed genes found in the microarray experiment. These dominant genes are then used to find out a dominant Genetic Network by applying the Bayesian Algorithm. Thus, the most dominant genetic network for the growth and development of the fungus under consideration was obtained. The genetic network represents the set of responsible genes in the growth process and their inter relationships. The Microarray data set represents the external manifestation of internal genetic activity resulting into genetic network. Here, from the set of 5289 genes in 47 consecutive time instances, were taken for analysis. Out of them, 400 most pertinent genes for the growth process were determined using a new technique namely ‘Fidelity Matrix Process’. Genetic Network for these 400 genes has been constructed and studied using Bayesian Belief Network Technique. The present reduction method was found to be more efficient in terms of computation when compared contemporary studies done many scientists. The results of the present study may be extensively applied in reducing a huge number of genetic expression rate data without any complex computation, studying unknown biological processes and systems, treating complicated diseases and even designing drugs for some incorrigible syndromes.

Genetic Network Responsible for the Growth of a Fungus Using the Concepts of Bayesian Algorithm.

Abstract: The Bayesian belief network is a powerful knowledge representation and reasoning tool under conditions of uncertainty to analyze gene expression