Multi Pattern Dynamic Time Warping For Automatic Speech Recognition

We are addressing the problem of jointly using multiple noisy speech patterns for automatic speech recognition (ASR), given that they come from the same class. If the user utters a word K times, the ASR system should try to use the information content in all the K patterns of the word simultaneously and improve its speech recognition accuracy compared to that of the single pattern based speech recognition. T address this problem, recently we proposed a Multi Pattern Dynamic Time Warping (MPDTW) algorithm to align the K patterns by finding the least distortion path between them. A Constrained Multi Pattern Viterbi algorithm was used on this aligned path for isolated word recognition (IWR). In this paper, we explore the possibility of using only the MPDTW algorithm for IWR. We also study the properties of the MPDTW algorithm. We show that using only 2 noisy test patterns (10 percent burst noise at -5 dB SNR) reduces the noisy speech recognition error rate by 37.66 percent when compared to the single pattern recognition using the Dynamic Time Warping algorithm.

Compact fiber Bragg grating dynamic strain sensor cum broadband thermometer for thermally unstable ambience

An instrument for simultaneous measurement of dynamic strain and temperature in a thermally unstable ambience has been proposed, based on fiber Bragg grating technology. The instrument can function as a compact and stand-alone broadband thermometer and a dynamic strain gauge. It employs a source wavelength tracking procedure for linear dependence of the output on the measurand, offering high dynamic range. Two schemes have been demonstrated with their relative merits. As a thermometer, the present instrumental configuration can offer a linear response in excess of 500 degrees C that can be easily extended by adding a suitable grating and source without any alteration in the procedure. Temperature sensitivity is about 0.06 degrees C for a bandwidth of 1 Hz. For the current grating, the upper limit of strain measurement is about 150 mu epsilon with a sensitivity of about 80 n epsilon Hz(-1/2). The major source of uncertainty associated with dynamic strain measurement is the laser source intensity noise, which is of broad spectral band. A low noise source device or the use of optical power regulators can offer improved performance. The total harmonic distortion is less than 0.5% up to about 50 mu epsilon, 1.2% at 100 mu epsilon and about 2.3% at 150 mu epsilon. Calibrated results of temperature and strain measurement with the instrument have been presented. Traces of ultrasound signals recorded by the system at 200 kHz, in an ambience of 100-200 degrees C temperature fluctuation, have been included. Also, the vibration spectrum and engine temperature of a running internal combustion engine has been recorded as a realistic application of the system.

A Kalman filter based strategy for linear structural system identification based on multiple static and dynamic test data

The problem of identification of stiffness, mass and damping properties of linear structural systems, based on multiple sets of measurement data originating from static and dynamic tests is considered. A strategy, within the framework of Kalman filter based dynamic state estimation, is proposed to tackle this problem. The static tests consists of measurement of response of the structure to slowly moving loads, and to static loads whose magnitude are varied incrementally; the dynamic tests involve measurement of a few elements of the frequency response function (FRF) matrix. These measurements are taken to be contaminated by additive Gaussian noise. An artificial independent variable τ, that simultaneously parameterizes the point of application of the moving load, the magnitude of the incrementally varied static load and the driving frequency in the FRFs, is introduced. The state vector is taken to consist of system parameters to be identified. The fact that these parameters are independent of the variable τ is taken to constitute the set of ‘process’ equations. The measurement equations are derived based on the mechanics of the problem and, quantities, such as displacements and/or strains, are taken to be measured. A recursive algorithm that employs a linearization strategy based on Neumann’s expansion of structural static and dynamic stiffness matrices, and, which provides posterior estimates of the mean and covariance of the unknown system parameters, is developed. The satisfactory performance of the proposed approach is illustrated by considering the problem of the identification of the dynamic properties of an inhomogeneous beam and the axial rigidities of members of a truss structure.

The stretching of single poly-ubiquitin molecules: Static versus dynamic disorder in the non-exponential kinetics of chain unfolding

Static disorder has recently been implicated in the non-exponential kinetics of the unfolding of single molecules of poly-ubiquitin under a constant force Kuo, Garcia-Manyes, Li, Barel, Lu, Berne, Urbakh, Klafter, and Fernandez, Proc. Natl. Acad. Sci. U. S. A. 107, 11336 (2010)]. In the present paper, it is suggested that dynamic disorder may provide a plausible, alternative description of the experimental observations. This suggestion is made on the basis of a model in which the barrier to chain unfolding is assumed to be modulated by a control parameter r that evolves in a parabolic potential under the action of fractional Gaussian noise according to a generalized Langevin equation. The treatment of dynamic disorder within this model is pursued using Zwanzig's indirect approach to noise averaging Acc. Chem. Res. 23, 148 (1990)]. In conjunction with a self-consistent closure scheme developed by Wilemski and Fixman J. Chem. Phys. 58, 4009 (1973); ibid. 60, 866 (1974)], this approach eventually leads to an expression for the chain unfolding probability that can be made to fit the corresponding experimental data very closely. (C) 2011 American Institute of Physics.

Autocalibration of dynamic bacterial growth model for water distribution system using GA

The presence of residual chlorine and organic matter govern the bacterial regrowth within a water distribution system. The bacterial growth model is essential to predict the spatial and temporal variation of all these substances throughout the system. The parameters governing the bacterial growth and biodegradable dissolved organic carbon (BDOC) utilization are difficult to determine by experimentation. In the present study, the estimation of these parameters is addressed by using simulation-optimization procedure. The optimal solution by genetic algorithm (GA) has indicated that the proper combination of parameter values are significant rather than correct individual values. The applicability of the model is illustrated using synthetic data generated by introducing noise in to the error-free measurements. The GA was found to be a potential tool in estimating the parameters controlling the bacterial growth and BDOC utilization. Further, the GA was also used for evaluating the sensitivity issues relating parameter values and objective function. It was observed that mu and k(cl) are more significant and dominating compared to the other parameters. But the magnitude of the parameters is also an important issue in deciding the dominance of a particular parameter. GA is found to be a useful tool in autocalibration of bacterial growth model and a sensitivity study of parameters.

A Gradient-Based Comparison Measure for Visual analysis of Multifield Data

We introduce a multifield comparison measure for scalar fields that helps in studying relations between them. The comparison measure is insensitive to noise in the scalar fields and to noise in their gradients. Further, it can be computed robustly and efficiently. Results from the visual analysis of various data sets from climate science and combustion applications demonstrate the effective use of the measure.

An Area-Efficient Noise-Adaptive Neural Amplifier in 130 nm CMOS Technology

Chronic recording of neural signals is indispensable in designing efficient brain–machine interfaces and to elucidate human neurophysiology. The advent of multichannel micro-electrode arrays has driven the need for electronics to record neural signals from many neurons. The dynamic range of the system can vary over time due to change in electrode–neuron distance and background noise. We propose a neural amplifier in UMC 130 nm, 1P8M complementary metal–oxide–semiconductor (CMOS) technology. It can be biased adaptively from 200 nA to 2 $mu{rm A}$, modulating input referred noise from 9.92 $mu{rm V}$ to 3.9 $mu{rm V}$. We also describe a low noise design technique which minimizes the noise contribution of the load circuitry. Optimum sizing of the input transistors minimizes the accentuation of the input referred noise of the amplifier and obviates the need of large input capacitance. The amplifier achieves a noise efficiency factor of 2.58. The amplifier can pass signal from 5 Hz to 7 kHz and the bandwidth of the amplifier can be tuned for rejecting low field potentials (LFP) and power line interference. The amplifier achieves a mid-band voltage gain of 37 dB. In vitro experiments are performed to validate the applicability of the neural low noise amplifier in neural recording systems.

Optimization of vibrating structures to reduce radiated noise

In this work, we propose an approach for reducing radiated noise from `light' fluid-loaded structures, such as, for example, vibrating structures in air. In this approach, we optimize the structure so as to minimize the dynamic compliance (defined as the input power) of the structure. We show that minimizing the dynamic compliance results in substantial reductions in the radiated sound power from the structure. The main advantage of this approach is that the redesign to minimize the dynamic compliance moves the natural frequencies of the structure away from the driving frequency thereby reducing the vibration levels of the structure, which in turn results in a reduction in the radiated sound power as an indirect benefit. Thus, the need for an acoustic and the associated sensitivity analysis is completely bypassed (although, in this work, we do carry out an acoustic analysis to demonstrate the reduction in sound power levels), making the strategy efficient compared to existing strategies in the literature which try to minimize some measure of noise directly. We show the effectiveness of the proposed approach by means of several examples involving both topology and stiffener optimization, for vibrating beam, plate and shell-type structures.

Effective contrast recovery in rapid dynamic near-infrared diffuse optical tomography using l(1)-norm-based linear image reconstruction method

Traditional image reconstruction methods in rapid dynamic diffuse optical tomography employ l(2)-norm-based regularization, which is known to remove the high-frequency components in the reconstructed images and make them appear smooth. The contrast recovery in these type of methods is typically dependent on the iterative nature of method employed, where the nonlinear iterative technique is known to perform better in comparison to linear techniques (noniterative) with a caveat that nonlinear techniques are computationally complex. Assuming that there is a linear dependency of solution between successive frames resulted in a linear inverse problem. This new framework with the combination of l(1)-norm based regularization can provide better robustness to noise and provide better contrast recovery compared to conventional l(2)-based techniques. Moreover, it is shown that the proposed l(1)-based technique is computationally efficient compared to its counterpart (l(2)-based one). The proposed framework requires a reasonably close estimate of the actual solution for the initial frame, and any suboptimal estimate leads to erroneous reconstruction results for the subsequent frames.

A low-noise low-power noise-adaptive neural amplifier in 0.13um CMOS technology

Chronic recording of neural signals is indispensable in designing efficient brain machine interfaces and in elucidating human neurophysiology. The advent of multichannel microelectrode arrays has driven the need for electronics to record neural signals from many neurons. The dynamic range of the system is limited by background system noise which varies over time. We propose a neural amplifier in UMC 130 nm, 2P8M CMOS technology. It can be biased adaptively from 200 nA to 2 uA, modulating input referred noise from 9.92 uV to 3.9 uV. We also describe a low noise design technique which minimizes the noise contribution of the load circuitry. The amplifier can pass signal from 5 Hz to 7 kHz while rejecting input DC offsets at electrode-electrolyte interface. The bandwidth of the amplifier can be tuned by the pseudo-resistor for selectively recording low field potentials (LFP) or extra cellular action potentials (EAP). The amplifier achieves a mid-band voltage gain of 37 dB and minimizes the attenuation of the signal from neuron to the gate of the input transistor. It is used in fully differential configuration to reject noise of bias circuitry and to achieve high PSRR.

Dynamic state estimation for identifying earthquake support motions in instrumented structures

The problem of identification of multi-component and (or) spatially varying earthquake support motions based on measured responses in instrumented structures is considered. The governing equations of motion are cast in the state space form and a time domain solution to the input identification problem is developed based on the Kalman and particle filtering methods. The method allows for noise in measured responses, imperfections in mathematical model for the structure, and possible nonlinear behavior of the structure. The unknown support motions are treated as hypothetical additional system states and a prior model for these motions are taken to be given in terms of white noise processes. For linear systems, the solution is developed within the Kalman filtering framework while, for nonlinear systems, the Monte Carlo simulation based particle filtering tools are employed. In the latter case, the question of controlling sampling variance based on the idea of Rao-Blackwellization is also explored. Illustrative examples include identification of multi-component and spatially varying support motions in linear/nonlinear structures.

High-resolution restoration of 3D structures from widefield images with extreme low signal-to-noise-ratio

Four-dimensional fluorescence microscopy-which records 3D image information as a function of time-provides an unbiased way of tracking dynamic behavior of subcellular components in living samples and capturing key events in complex macromolecular processes. Unfortunately, the combination of phototoxicity and photobleaching can severely limit the density or duration of sampling, thereby limiting the biological information that can be obtained. Although widefield microscopy provides a very light-efficient way of imaging, obtaining high-quality reconstructions requires deconvolution to remove optical aberrations. Unfortunately, most deconvolution methods perform very poorly at low signal-to-noise ratios, thereby requiring moderate photon doses to obtain acceptable resolution. We present a unique deconvolution method that combines an entropy-based regularization function with kernels that can exploit general spatial characteristics of the fluorescence image to push the required dose to extreme low levels, resulting in an enabling technology for high-resolution in vivo biological imaging.

Online Estimation of Evolving Human Visual Interest

Regions in video streams attracting human interest contribute significantly to human understanding of the video. Being able to predict salient and informative Regions of Interest (ROIs) through a sequence of eye movements is a challenging problem. Applications such as content-aware retargeting of videos to different aspect ratios while preserving informative regions and smart insertion of dialog (closed-caption text) into the video stream can significantly be improved using the predicted ROIs. We propose an interactive human-in-the-loop framework to model eye movements and predict visual saliency into yet-unseen frames. Eye tracking and video content are used to model visual attention in a manner that accounts for important eye-gaze characteristics such as temporal discontinuities due to sudden eye movements, noise, and behavioral artifacts. A novel statistical-and algorithm-based method gaze buffering is proposed for eye-gaze analysis and its fusion with content-based features. Our robust saliency prediction is instantiated for two challenging and exciting applications. The first application alters video aspect ratios on-the-fly using content-aware video retargeting, thus making them suitable for a variety of display sizes. The second application dynamically localizes active speakers and places dialog captions on-the-fly in the video stream. Our method ensures that dialogs are faithful to active speaker locations and do not interfere with salient content in the video stream. Our framework naturally accommodates personalisation of the application to suit biases and preferences of individual users.

Bayesian parameter identification in dynamic state space models using modified measurement equations

When Markov chain Monte Carlo (MCMC) samplers are used in problems of system parameter identification, one would face computational difficulties in dealing with large amount of measurement data and (or) low levels of measurement noise. Such exigencies are likely to occur in problems of parameter identification in dynamical systems when amount of vibratory measurement data and number of parameters to be identified could be large. In such cases, the posterior probability density function of the system parameters tends to have regions of narrow supports and a finite length MCMC chain is unlikely to cover pertinent regions. The present study proposes strategies based on modification of measurement equations and subsequent corrections, to alleviate this difficulty. This involves artificial enhancement of measurement noise, assimilation of transformed packets of measurements, and a global iteration strategy to improve the choice of prior models. Illustrative examples cover laboratory studies on a time variant dynamical system and a bending-torsion coupled, geometrically non-linear building frame under earthquake support motions. (C) 2015 Elsevier Ltd. All rights reserved.

Acoustic Noise Characterization of Space-Vector Modulated Induction Motor Drives-An Experimental Approach

This paper presents an experimental procedure to determine the acoustic and vibration behavior of an inverter-fed induction motor based on measurements of the current spectrum, acoustic noise spectrum, overall noise in dB, and overall A-weighted noise in dBA. Measurements are carried out on space-vector modulated 8-hp and 3-hp induction motor drives over a range of carrier frequencies at different modulation frequencies. The experimental data help to distinguish between regions of high and low acoustic noise levels. The measurements also bring out the impact of carrier frequency on the acoustic noise. The sensitivity of the overall noise to carrier frequency is indicative of the relative dominance of the high-frequency electromagnetic noise over mechanical and aerodynamic components of noise. Based on the measured current and acoustic noise spectra, the ratio of dynamic deflection on the stator surface to the product of fundamental and harmonic current amplitudes is obtained at each operating point. The variation of this ratio of deflection to current product with carrier frequency indicates the resonant frequency clearly and also gives a measure of the amplification of vibration at frequencies close to the resonant frequency. This ratio is useful to predict the magnitude of acoustic noise corresponding to significant time-harmonic currents flowing in the stator winding.