Input Parameter Vector Research Articles

Диагностирование газовоздушного тракта тепловозного дизеля с использованием интеллектуального классификатора

Purpose: Efficiency increase of operational monitoring of technical state of gas-air tract (GAT) of a locomotive diesel, assessment of possibility of using locomotive regular measuring tools to monitor diesel GAT technical state; feasibility justification for expanding the list of GAT parameters that are monitored by locomotive board diagnostics. Methods: An intelligent classifier based on Support Vector Machine (SVM) algorithm is used to determine the current technical condition (TC) class of GAT. Training sample of 96 item volume for eight classes of GAT TC was formed using “Diesel-RK” software complex. Classifier structure three variants were studied, differing in the set of input parameter vector components: regular, with two inputs (pressure increase degree in a supercharger and exhaust gas temperature before a turbine), expanded, with additional control of turbocharger rotor rotation speed, and proposed, with control of air instantaneous consumption by a diesel instead. Model quality was assessed by cross-validation on five fragments and by classification results for test sample which wasn’t used while training. Results: Classification accuracy of testing sample by classifier with input parameter regular set doesn’t exceed 41%. The expansion of being controlled list of parameter on account of inclusion therein of turbocharger rotor rotation speed does not noticeably affect diagnostic system efficiency, increasing diagnostic accuracy only till 43%. At the same time, rotor rotation speed replacement with air instantaneous consumption rises testing sample classification accuracy till 91%. Practical significance: The effectiveness of the use of machine learning methods for operative monitoring of diesel GAT technical state at condition of optimization of the list of diesel parameters, controlled by locomotive board diagnostics tools, is confirmed. With the purpose to increase control efficiency it is recommended to include diesel instantaneous air consumption into the list with parallel exclusion from their of a turbocharger rotor rotation frequency.

Text-driven Speech Animation with Emotion Control

In this paper, we present a new approach to creating speech animation with emotional expressions using a small set of example models. To generate realistic facial animation, two example models called key visemes and expressions are used for lip-synchronization and facial expressions, respectively. The key visemes represent lip shapes of phonemes such as vowels and consonants while the key expressions represent basic emotions of a face. Our approach utilizes a text-to-speech (TTS) system to create a phonetic transcript for the speech animation. Based on a phonetic transcript, a sequence of speech animation is synthesized by interpolating the corresponding sequence of key visemes. Using an input parameter vector, the key expressions are blended by a method of scattered data interpolation. During the synthesizing process, an importance-based scheme is introduced to combine both lip-synchronization and facial expressions into one animation sequence in real time (over 120Hz). The proposed approach can be applied to diverse types of digital content and applications that use facial animation with high accuracy (over 90%) in speech recognition.

Effect of Linearization in a WNT Signaling Model.

A nonlinear model consisting of a system of coupled ordinary differential equations (ODE), describing a biological process linked with cancer development, is linearized using Taylor series and tested against different magnitudes of input perturbations, in order to investigate the extent to which the linearization is accurate. The canonical wingless/integrated (WNT) signaling pathway is considered. The linearization procedure is described, and special considerations for linearization validity are analyzed. The analytical properties of nonlinear and linearized systems are studied, including aspects such as existence of steady state and initial value sensitivity. Linearization is a useful tool for speeding up drug response computations or for providing analytical answers to problems such as required drug concentrations. A Monte Carlo-based error testing workflow is employed to study the errors introduced by the linearization for different input conditions and parameter vectors. The deviations between the nonlinear and the linearized system were found to increase in a polynomial fashion w.r.t. the magnitude of tested perturbations. The linearized system closely followed the original one for perturbations of magnitude within 10% of the base input vector which yielded the state-space fixed point used for the linearization.

Expert system based on computer vision to estimate the content of impurities in olive oil samples

The determination of the content of impurities is a very frequent analysis performed on virgin olive oil samples, but the official method is quite work-intensive, and it would be convenient to have an alternative approximate method to evaluate the performance of the impurity removal process. In this work we develop a system based on computer vision and pattern recognition to classify the content of impurities of the olive oil samples in three sets, indicative of the goodness of the separation process of olive oil after its extraction from the paste. Starting from the histograms of the channels of the Red–Green–Blue (RGB), CIELAB and Hue-Saturation-Value (HSV) color spaces, we construct an initial input parameter vector and perform a feature extraction previous to the classification. Several linear and non-linear feature extraction techniques were evaluated, and the classifiers used were Support Vector Machines (SVMs) and Artificial Neural Networks (ANNs). The best classification rate achieved was 87.66%, obtained using Kernel Principal Components Analysis (KPCA) and a grade-3-polynomial kernel SVM. The best result using ANNs was 82.38%, yielded by the use of Principal Component Analysis (PCA) with the Perceptron.

CHARACTERIZING THE FORMATION HISTORY OF MILKY WAY LIKE STELLAR HALOS WITH MODEL EMULATORS

We use the semi-analytic model ChemTreeN, coupled to cosmological N-body simulations, to explore how different galaxy formation histories can affect observational properties of Milky Way-like galaxies' stellar haloes and their satellite populations. Gaussian processes are used to generate model emulators that allow one to statistically estimate a desired set of model outputs at any location of a p-dimensional input parameter space. This enables one to explore the full input parameter space orders of magnitude faster than could be done otherwise. Using mock observational data sets generated by ChemTreeN itself, we show that it is possible to successfully recover the input parameter vectors used to generate the mock observables if the merger history of the host halo is known. However, our results indicate that for a given observational data set the determination of "best fit" parameters is highly susceptible to the particular merger history of the host. Very different halo merger histories can reproduce the same observational dataset, if the "best fit" parameters are allowed to vary from history to history. Thus, attempts to characterize the formation history of the Milky Way using these kind of techniques must be performed statistically, analyzing large samples of high resolution N-body simulations.

Sign Language to Number by Neural Network

this paper is presented an automatic deaf language to number recognition system. Sign language number recognition system lays down foundation for hand shape recognition which addresses real and current problems in signing in the deaf community and leads to practical applications. The scheme is based on neural network (NN) classifier using a back propagation. The input for the sign language number recognition system is 1000 Indian Sign Language number images with 640 x 480 pixels size. The input parameter vector to neural network is the Fisher score, which represents the derivate of the matrix of symbol probability in hidden Markov model (HMM). The HMM, which needs a sequence to be trained and used, is fed by the hand contour chain code. Besides, an improvement on the calculation of Fisher score is introduced by means of reducing the kernel scores variance. The error ratio classifying hand number of the proposed system is less than 1.4% with our database. The system learns and recognizes the Indian Sign Language number in training and testing phase using Hidden Markov Model and neural network. The system uses neural network for training and testing phase. The sign language recognizer could recognize Indian sign language number with 98.52% accuracy. Keywordsvision, Human computer interaction, Sign language, Image processing, Chain Code, HMM, Fisher score, Neural network, Sign language recognition.

Practical speaker‐independent voice recognition using segmental features

AbstractThis paper reports a practical method that achieves speaker‐independent large‐vocabulary voice recognition with high accuracy and high noise immunity but with small computational complexity. The first feature of the method is that highly accurate recognition is realized by using an acoustic model in which the input consists of the segmental features formed by the analysis parameters of multiple frames. The second feature is that the likelihood corresponding to the output probability of each state of the acoustic model is calculated by a linear expression with the input parameter vector as the variable. The linear expression is derived from the equal‐covariance assumption. This linear expression reduces the computational complexity and the required memory capacity even if segmental features are used, without degrading recognition performance. The third feature is that a stable word spotting function detects correct solution candidates from the noise interval by including the idea of the a posteriori probability in the likelihood calculation. This word spotting function allows voice recognition which is robust to noise to be realized. The effectiveness of the proposed system was demonstrated in a recognition experiment with superimposed interior noise of a running vehicle. © 2004 Wiley Periodicals, Inc. Electron Comm Jpn Pt 2, 87(2): 73–81, 2004; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/ecjb.10217

Feature extraction and dimensionality reduction algorithms and their applications in vowel recognition

Feature extraction is an important component of a pattern recognition system. It performs two tasks: transforming input parameter vector into a feature vector and/or reducing its dimensionality. A well-defined feature extraction algorithm makes the classification process more effective and efficient. Two popular methods for feature extraction are linear discriminant analysis (LDA) and principal component analysis (PCA). In this paper, the minimum classification error (MCE) training algorithm (which was originally proposed for optimizing classifiers) is investigated for feature extraction. A generalized MCE (GMCE) training algorithm is proposed to mend the shortcomings of the MCE training algorithm. LDA, PCA, and MCE and GMCE algorithms extract features through linear transformation. Support vector machine (SVM) is a recently developed pattern classification algorithm, which uses non-linear kernel functions to achieve non-linear decision boundaries in the parametric space. In this paper, SVM is also investigated and compared to linear feature extraction algorithms.

Global adaptive output feedback stabilization of nonlinear systems of any relative degree with unknown high-frequency gains

This paper presents an algorithm for adaptive output feedback stabilization of nonlinear single-input/single-output, minimum phase systems of any relative degree which are linear with respect to input and unknown constant parameter vector. The algorithm does not require the knowledge of the sign of the high-frequency gain.

A Scaled Stochastic Approximation Algorithm

Consider a stochastic system of such complexity that its performance can only be evaluated by using simulation or direct experimentation. To optimize the expected performance of such systems as a function of several continuous input parameters (decision variables), we present a “scaled” stochastic approximation algorithm for finding the zero (root) of the gradient of the response function. In each iteration of the scaled algorithm, two independent gradient estimates are sampled at the current estimate of the optimal input-parameter vector to compute a scale-free estimate of the next search direction. We establish sufficient conditions to ensure strong consistency and asymptotic normality of the resulting estimator of the optimal input-parameter vector. Strong consistency is also established for a variant of the scaled algorithm with Kesten's acceleration. An experimental performance comparison of the scaled algorithm and the classical Robbins-Monro algorithm in two simple queueing systems reveals some of the practical advantages of the scaled algorithm.