Ching-Fa Chen

Stability Analysis for a Class of Uncertain Discrete Singularly Perturbed Systems With Multiple Time Delays

In this paper, the robust stability problem for a class of nominally stable uncertain discrete singularly perturbed linear systems with multiple time delays is considered. A stability criterion for the slow and fast subsystems is first derived. A delay-dependent criterion is then proposed to guarantee the robust stability of the system subject to norm-bounded perturbations. A numerical example is provided to illustrate our main results.

Speech recognition via Hidden Markov Model and neural network trained by genetic algorithm

It is the goal of this paper to find a more suitable architecture for speech recognition to be implemented on a chip. This paper uses the Hidden Markov Model (HMM) and the Artificial Neural Networks (ANN) for speech recognition. The speech recognition algorithms are then implemented on the Field Programmable Gate Array (FPGA) chip for a comparison of speech recognition speed on hardware for HMM and ANN. In order to obtain a solution more close to the optimal solution for the parameters of ANN, this paper use genetic algorithm (GA) to train the ANN. It will be seen that the ANN trained by GA will get a better performance than that trained by gradient-descent method.

The robust D-stability analysis of uncertain discrete-delay descriptor systems via genetic algorithms

In this paper, work concerning robust D-stability analysis is considered for discrete descriptor systems with structured uncertainties and multiple time delays. A genetic algorithm is used to make the stability test of the main result more efficient. A delay-dependent robust D-stability criterion is proposed to ensure that all poles of the discrete multiple time-delay descriptor systems subject to uncertainties are located within a disk contained in the unit circle. It will be seen that the system considered in this paper is more general and the proposed robust D-stability criterion is less conservative and easier to check than those presented in recent research. Moreover, a delay-dependent criterion is derived to guarantee that the system is regular, impulse-free, and D-stable. Since the boundary test occurs during the check of the criteria, a genetic algorithm is used to make it more precise. Finally, an example is introduced to illustrate our main results.

Stability Analysis for a Class of Singularly Perturbed Systems With Multiple Time Delays

The paper is to investigate the asymptotic stability for a general class of linear timeinvariant singularly perturbed systems with multiple non-commensurate time delays. It is a common practice to investigate the asymptotic stability of the original system by establishing that of its slow subsystem and fast subsystem. A frequency-domain approach is first presented to determine a sufficient condition for the asymptotic stability of the slow subsystem (reduced-order model), which is a singular system with multiple time delays, and the fast subsystem. Two delay-dependent criteria,«-dependent and«-independent, are then proposed in terms of the H‘-norm for the asymptotic stability of the original system. Furthermore, a simple estimate of an upper bound«* of singular perturbation parameter « is proposed so that the original system is asymptotically stable for any«P~0,«*! .T wo numerical examples are provided to illustrate the use of our main results. @DOI: 10.1115/1.1793172#

Performances comparison between Improved DHMM and Gaussian Mixture HMM for speech recognition

This paper compares the performances, recognition rate and computation speed, between an Improved Discrete Hidden Markov Model (DHMM) and Gaussian Mixture Hidden Markov Model (GMHMM) for Mandarin speech recognition. The fuzzy vector quantization (FVQ) is used to improve the modeling of DHMM for the speech recognition. A codebook for DHMM will be first trained by K-means algorithms using Mandarin training speech feature. Then, based on the trained codebook, the speech features are quantized by the fuzzy sets and then are statistically applied to train the model of DHMM. Experimental results in this paper will show that the speech recognition rate can be improved by using FVQ algorithm to train the model of DHMM. The recognition rate by using an improved DHMM is only a little bit less than that by using GMHMM. However, the computation time for speech recognition by using improved DHMM is much less than that by using GMHMM. These results reveal that the improved DHMM is more suitable to real-time applications than GMHMM.

Genetic algorithm on speech recognition by using DHMM

This paper uses genetic algorithms to train a codebook for the modeling of Discrete Hidden Markov Model (DHMM) applied to speech recognition. The GA-trained DHMM is then used to increase the recognition rate for Mandarin speeches. Vector quantization based on a codebook is a fundamental process to recognize the speech signal by DHMM. A codebook will be first trained by genetic algorithms through Mandarin speech features. The speech features are then quantized based on the trained codebook. Subsequently, the quantized speech features are statistically used to train the model of DHMM for speech recognition. All the speech features to be recognized should go through the codebook before being fed into the DHMM model for recognition. Experimental results show that the speech recognition rate can be improved by using genetic algorithms to train the model of DHMM.

Genetic Algorithm and Time-Scale Separation on the Reduced-Order Observer-Based Control Design for a Class of Discrete Systems

The design of the control for discrete multiple time-delay systems subject to input constraint is considered in this paper. Genetic algorithms will be applied to adjust the control gain and then a observer-based feedback control will be designed to stabilize the closed-loop system with input constraint. Moreover, using the two time-scale property of the system, the slow and fast subsystem of the discrete systems will be derived. Then the controls are design for the two subsystems such that the two subsystems are both stabile. The control of the original full-order system is then derived from the two controls. Finally, an application example will also be given in this paper to illustrate the results of this paper.

Genetic algorithm on fuzzy codebook training for speech recognition

A genetic algorithm is used to train the fuzzy membership function of a fuzzy codebook for the modeling of Discrete Hidden Markov Model (DHMM) applied to Mandarin speech recognition. Vector quantization for a speech feature based on a codebook is a fundamental process to recognize the speech signal by DHMM. A codebook with fuzzy membership functions corresponding to each vector in the codebook will be first trained by genetic algorithms (GAs) through speech features. The trained fuzzy codebook is then used to quantize the speech features. Subsequently, the quantized speech statistical features are used to model the DHMM for each speech. Besides, all the speech features to be recognized will go through the fuzzy codebook for quantization before being fed into the DHMM model for recognition. Experimental results show that both the speech recognition rate and computation time for recognition can be improved by the proposed strategy.

Robust stability for a class of two-time-scale time-delay neutral systems

In this paper, the robust stability problem for a class of uncertain two-time-scale neutral systems subject to unstructured perturbations is investigated. The properties of H/sub /spl infin//-norm are used throughout this paper to derive the robust stability criterion. A frequency-domain delay-dependent criterion for the asymptotic stability of the slow neutral subsystem and the fast subsystem of the nominal two-time-scale neutral system is first presented. Under the condition that the slow neutral and fast subsystems of the nominal system are both asymptotically stable, we then propose the allowable bounds for the system uncertainties such that the slow neutral subsystem and the fast subsystem of the original uncertain two-time scale neutral system are both asymptotically stable. Finally, a stability bound /spl epsi//sup */ of the small parameter /spl epsi/ is given such that the original uncertain two-time scale neutral system is asymptotically stable for any /spl epsi/ /spl isin/ (0, /spl epsi//sup */).