• ETRI Journal
• /
• v.34 no.5
• /
• pp.696-705
• /
• 2012

Accurate synthesis models based on artificial neural networks (ANNs) are proposed to directly obtain the physical dimensions of an asymmetric coplanar waveguide with conductor backing and substrate overlaying (ACPWCBSO). First, the ACPWCBSO is analyzed with the conformal mapping technique (CMT) to obtain the training data. Then, a modified genetic-algorithm-Levenberg-Marquardt (GA-LM) algorithm is adopted to train ANNs. In the algorithm, the maximal relative error (MRE) is used as the fitness function of the chromosomes to guarantee that the MRE is small, while the mean square error is used as the error function in LM training to ensure that the average relative error is small. The MRE of ANNs trained with the modified GA-LM algorithm is less than 8.1%, which is smaller than those trained with the existing GA-LM algorithm and the LM algorithm (greater than 15%). Lastly, the ANN synthesis models are validated by the CMT analysis, electromagnetic simulation, and measurements.

• MALSORI
• /
• v.68
• /
• pp.65-74
• /
• 2008

A statistical parametric speech synthesis system based on the hidden Markov models (HMMs) has grown in popularity over the last few years, because it needs less memory and low computation complexity and is suitable for the embedded system in comparison with a corpus-based unit concatenation text-to-speech (TTS) system. It also has the advantage that voice characteristics of the synthetic speech can be modified easily by transforming HMM parameters appropriately. In this paper, we present experimental results of voice characteristics conversion using the HMM-based Korean speech synthesis system. The results have shown that conversion of voice characteristics could be achieved using a few sentences uttered by a target speaker. Synthetic speech generated from adapted models with only ten sentences was very close to that from the speaker dependent models trained using 646 sentences.

• Journal of the Korea Institute of Military Science and Technology
• /
• v.4 no.1
• /
• pp.93-103
• /
• 2001

A reflection signal in an active sonar using a high frequency is mainly formed of a specular reflection from the surface of an object along with several equivalent scatters inside, which are characterized by the spatial distribution of the highlight on the object. This study analyze the existing echo signal synthesis models eq, random distribution model, equivalent interval distribution model & MUTAHID(Modified Underwater TArget by HIlight Distribution) model for simulated target, and compare the characteristics of the reflected signal synthesis results for each model in various conditions. These highlight distribution models can be efficiently applied to the simulated target signals synthesis of various real systems requiring the echo signal synthesis on the underwater target.

• Speech Sciences
• /
• v.9 no.1
• /
• pp.207-215
• /
• 2002

Perceptual evaluation of duration models of spoken Korean was carried out based on the Classification and Regression Tree (CART) model for text-to-speech conversion. A reference set of durations was produced by a commercial text-to-speech synthesis system for comparison. The duration model which was built in the previous research (Chung & Huckvale, 2001) was applied to a Korean language speech synthesis diphone database, 'Hanmal (HN 1.0)'. The synthetic speech produced by the CART duration model was preferred in the subjective preference test by a small margin and the synthetic speech from the commercial system was superior in the clarity test. In the course of preparing the experiment, a labeled database of spoken Korean with 670 sentences was constructed. As a result of the experiment, a trained duration model for speech synthesis was obtained. The 'Hanmal' diphone database for Korean speech synthesis was also developed as a by-product of the perceptual evaluation.

• Journal of Institute of Control, Robotics and Systems
• /
• v.10 no.5
• /
• pp.450-457
• /
• 2004

In this research, the automatic synthesis of knowledge models is proposed. which are the basis of the methods using qualitative models adapted widely in fault diagnosis and hazard evaluation of chemical processes. To provide an easy and fast way to construct accurate causal model of the target process, the Role-Behavior modeling method is developed to represent the knowledge of modularized process units. In this modeling method, Fault-Behavior model and Structure-Role model present the relationship of the internal behaviors and faults in the process units and the relationship between process units respectively. Through the multiple modeling techniques, the knowledge is separated into what is independent of process and dependent on process to provide the extensibility and portability in model building, and possibility in the automatic synthesis. By taking advantage of the Role-Behavior Model, an algorithm is proposed to synthesize the plant-wide causal model, Fault-Causality Graph (FCG) from specific Fault-Behavior models of the each unit process, which are derived from generic Fault-Behavior models and Structure-Role model. To validate the proposed modeling method and algorithm, a system for building FCG model is developed on G2, an expert system development tool. Case study such as CSTR with recycle using the developed system showed that the proposed method and algorithm were remarkably effective in synthesizing the causal knowledge models for diagnosis of chemical processes.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.15 no.5
• /
• pp.1590-1609
• /
• 2021

Laughter is one of the most important nonverbal sound that human generates. It is a means for expressing his emotions. The acoustic and contextual features of this specific sound are different from those of speech and many difficulties arise during their modeling process. During this work, we propose an audio laughter generation system based on unsupervised generative models: the autoencoder (AE) and its variants. This procedure is the association of three main sub-process, (1) the analysis which consist of extracting the log magnitude spectrogram from the laughter database, (2) the generative models training, (3) the synthesis stage which incorporate the involvement of an intermediate mechanism: the vocoder. To improve the synthesis quality, we suggest two hybrid models (LSTM-VAE, GRU-VAE and CNN-VAE) that combine the representation learning capacity of variational autoencoder (VAE) with the temporal modelling ability of a long short-term memory RNN (LSTM) and the CNN ability to learn invariant features. To figure out the performance of our proposed audio laughter generation process, objective evaluation (RMSE) and a perceptual audio quality test (listening test) were conducted. According to these evaluation metrics, we can show that the GRU-VAE outperforms the other VAE models.

• Journal of Multimedia Information System
• /
• v.2 no.3
• /
• pp.263-274
• /
• 2015

The subject of this work is the complex discrete systems simulation special features with the aid of dynamic graph models. The proposed simulation technique allows to determine the ways for tasks solutions in terms of discrete systems analysis and synthesis of various complication: one-dimensional and multidimensional, steady and unstable, with the pulse elements abnormal operating mode and others. Often complex control systems analysis and synthesis task solutions, via classical approach comes out to be insolvent, because of the computational problems. The application of graph models allows to perform clear and strict characterization and computer procedures automation. The optimal controls synthesis algorithm presented in this paper, transferring the discrete system from target initial state to target final state within the minimum time, allows to consider the zero initial conditions systems, with the initial potential energy, with the control actions limitations and complex pulse elements operating mode.

• Proceedings of the Korean Society for Noise and Vibration Engineering Conference
• /
• 2002.05a
• /
• pp.651-659
• /
• 2002

Component Mode Synthesis (CMS) is a dynamic substructuring technique to get an approximate eigensolutions of large degree-of-freedom structures divisible into several components. But, In practice. most of large structures are modeled by different teams of engineers. and their respective finite element models often require different mesh resolutions. As a result, the finite element substructure models can be non-conforming and/or incompatible. In this work, A hybrid version of component mode synthesis using a localized lagrange multiplier to treat the non-conforming mesh problem was derived. Evolution Strategies (ESs) is a stochastic numerical optimization technique and has shown a robust performance for solving deterministic problems. An ESs conducts its search by processing a population of solutions for an optimization problem based on principles from natural evolution. An optimization example for raising the first natural frequency of a plate structure using beam stiffeners was presented using hybrid component mode synthesis and robust evolution strategies (RES) optimization technique. In the example. the design variables are the positions and lengths of beam stiffeners.

• Journal of The Korean Astronomical Society
• /
• v.29 no.spc1
• /
• pp.49-51
• /
• 1996

New population synthesis models, with the effects of metallicity spread and the horizontal-branch (HB) morphology, provide a way to break the well-known age-metallicity degeneracy in the analysis of the integrated light of elliptical galaxies. Our models suggest that the far- UV radiation of these systems is dominated by a minority population of metal-poor, hot HB stars and their post-HB progeny, while the optical radiation is dominated by a metal-rich population. The systematic variation of UV upturn depends on the contribution from metal-poor, hot HB stars and their post-HB progeny, which in turn depends on the ages of old stellar populations in galaxies. Our result implies a prolonged epoch of galaxy formation, in the sense that more massive galaxies (in denser environments) formed first. Our models also suggest that the strenghth of H$\beta$ index is strongly affected by HB stars, and hence previous age estimation without detailed modeling of the HB would underestimate the ages of ellipticals by $\~$7 Gyr.

• Transactions of the Korean Society of Mechanical Engineers A
• /
• v.25 no.3
• /
• pp.495-501
• /
• 2001

In this paper, a new technique which models the joints characteristics through reduction of DOFs of structures with joints using component mode synthesis (CMS) method is proposed. Bolt joints are modeled by mass-spring systems. Also generalized mass and stiffness matrices for this models are introduced. Because bolt joints have influence on eigenvalues of structures, exact eigenvalues from modal test are used. The results show that the behaviors of structures with bolt joints depend to a large extent on the translational DOFs and not on rotational DOFs of mass and stiffness matrices of bolts. Furthermore it is confirmed that lumped mass-spring systems as models of bolt joints are effective models considering the facts that joint characteristics converged to constant values in some iterations and eignevalues from proposed method are in good agreement with ones from modal test.