Browsing by Subject "signal processing"
Now showing 1 - 4 of 4
Results Per Page
Sort Options
Item Development of an infrared absorption spectroscope based on linear variable filters(2009-05-15) Nogueira, Felipe GuimaraesThe objective of this thesis is to develop a low-cost infrared absorption spectroscope based on linear variable filter (LVF) technology for the automated detection of gases and vapors, and the semi-automated detection of liquids. This instrument represents an alternative to electronic-nose instruments based on cross-selective gas sensor arrays. Instead, the proposed instrument uses the idea of computational ?pseudo-sensors?, whereby spectral lines in an analytical instrument are clustered into groups and used as independent variables. We characterize the system on a number of performance metrics, uncovering its detection limits and resolving power. We present calibration methods to estimate the concentration of analytes in a matrix of absorbing species, as well as signal processing techniques for spectral classification. Specifically, we validate the instrument on a mixture calibration problem with simple and complex chemicals, and compare the efficiency of different calibration methods to estimate the concentration of one analyte in the matrix. Moreover, we demonstrate the use of the instrument on two real-world applications in the foodstuffs domain: oil adulteration and trans fatty acid (TFA) detection. The instrument, combined with signal processing techniques, is able to fully discriminate oils, as well as classify margarine and spreads onto high-TFA and low-TFA groups. Despite operating at a low spectral resolution, our results show that the LVF based spectroscope is a promising alternative to traditional analytical techniques for selected niche applications.Item Genomic applications of statistical signal processing(2009-05-15) Zhao, WentaoBiological phenomena in the cells can be explained in terms of the interactions among biological macro-molecules, e.g., DNAs, RNAs and proteins. These interactions can be modeled by genetic regulatory networks (GRNs). This dissertation proposes to reverse engineering the GRNs based on heterogeneous biological data sets, including time-series and time-independent gene expressions, Chromatin ImmunoPrecipatation (ChIP) data, gene sequence and motifs and other possible sources of knowledge. The objective of this research is to propose novel computational methods to catch pace with the fast evolving biological databases. Signal processing techniques are exploited to develop computationally efficient, accurate and robust algorithms, which deal individually or collectively with various data sets. Methods of power spectral density estimation are discussed to identify genes participating in various biological processes. Information theoretic methods are applied for non-parametric inference. Bayesian methods are adopted to incorporate several sources with prior knowledge. This work aims to construct an inference system which takes into account different sources of information such that the absence of some components will not interfere with the rest of the system. It has been verified that the proposed algorithms achieve better inference accuracy and higher computational efficiency compared with other state-of-the-art schemes, e.g. REVEAL, ARACNE, Bayesian Networks and Relevance Networks, at presence of artificial time series and steady state microarray measurements. The proposed algorithms are especially appealing when the the sample size is small. Besides, they are able to integrate multiple heterogeneous data sources, e.g. ChIP and sequence data, so that a unified GRN can be inferred. The analysis of biological literature and in silico experiments on real data sets for fruit fly, yeast and human have corroborated part of the inferred GRN. The research has also produced a set of potential control targets for designing gene therapy strategies.Item Improving Fluorescence Lifetime Imaging Microscopy Deconvolution Using Constrained Laguerre Basis Functions(2014-04-25) Khatkhatay, Mohammed MFluorescence lifetime imaging microscopy (FLIM) is a noninvasive invasive optical imaging modality which is finding new applications in medical imaging. In FLIM, the fluorescence time decay is measured at a pixel. The fluorescence impulse response function (IRF) is then estimated using a deconvolution of the instrument response and the measured fluorescence time decay. Two of the challenges facing FLIM are speed of the deconvolution and the accuracy of the IRFs. The linear expansion of the fluorescence decays based on the orthonormal Laguerre basis functions (LBFs) is among the fastest methods for estimating the IRFs. The automated implementation to optimize the Laguerre parameter improves the speed of the deconvolution using the LBFs but uses a global optimization. Therefore, the IRFs do not necessarily mimic exponential time decays, or monotonically decreasing functions. On the other hand, applying a constraint to the LBFs using the Active Set Nonnegative Least Squares (NNLS) method improves the IRF estimation. The estimation of the Laguerre parameter using the NNLS method, however, is about 10-15x slower. By combining these two deconvolution techniques, we found that the deconvolution time is similar to the automated global Laguerre parameter deconvolution while the IRF estimation always results in a monotonically decreasing function.Item Robust Framework for System Architecture and Hand-offs in Wireless and Cellular Communication Systems(2010-01-14) Varma, Vishal V.Robustness of a system has been defined in various ways and a lot of work has been done to model the robustness of a system, but quantifying or measuring robustness has always been very difficult. In this research, we develop a framework for robust system architecture. We consider a system of a linear estimator (multiple tap filter) and then attempt to model the system performance and robustness in a graphical manner, which admits an analysis using the differential geometric concepts. We compare two different perturbation models, namely the gradient with biased perturbations (sub-optimal model) of a surface and the gradient with unbiased perturbations (optimal model), and observe the values to see which of them can alternately be used in the process of understanding or measuring robustness. In this process we have worked on different examples and conducted many simulations to find if there is any consistency in the two models. We propose the study of robustness measures for estimation/prediction in stationary and non-stationary environment using differential geometric tools in conjunction with probability density analysis. Our approach shows that the gradient can be viewed as a random variable and therefore used to generate probability densities, allowing one to draw conclusions regarding the robust- ness. As an example, one can apply the geometric methodology to the prediction of time varying deterministic data in imperfectly known non-stationary distribution. We also compare stationary to non-stationary distribution and prove that robustness is reduced by admitting residual non-stationarity. We then research and develop a robust iterative handoff algorithm, relating generally to methods, devices and systems for reselecting and then handing over a mobile communications device from a first cell to a second cell in a cellular wireless communications system (GPRS, W-CDMA or OFDMA). This algorithm results in significant decrease in amount of power and/or result is a decrease of break in communications during an established voice call or other connection, in the field, thereby outperforming prior art.