• The Journal of the Acoustical Society of Korea
• /
• v.22 no.7
• /
• pp.603-611
• /
• 2003

This paper proposes an effective feature compensation scheme based on speech model for achieving robust speech recognition. Conventional model-based method requires off-line training with noisy speech database and is not suitable for online adaptation. In the proposed scheme, we can relax the off-line training with noisy speech database by employing the parallel model combination technique for estimation of correction factors. Applying the model combination process over to the mixture model alone as opposed to entire HMM makes the online model combination possible. Exploiting the availability of noise model from off-line sources, we accomplish the online adaptation via MAP (Maximum A Posteriori) estimation. In addition, the online channel estimation procedure is induced within the proposed framework. For more efficient implementation, we propose a selective model combination which leads to reduction or the computational complexities. The representative experimental results indicate that the suggested algorithm is effective in realizing robust speech recognition under the combined adverse conditions of additive background noise and channel distortion.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.44 no.2 s.314
• /
• pp.93-101
• /
• 2007

To jointly optimize the spatial registration and the exposure compensation, an iterative registration algorithm, the Lucas-Kanade algorithm, is combined with an exposure compensation algorithm, which is based on the histogram transformation function. Based on a simple regression model, a nonparametric estimator, the empirical conditional mean, and its polynomial fitting are used as histogram transformation functions for the exposure compensation. Since the proposed algorithm is composed of separable optimization phases, the proposed algorithm is more advantageous than the joint approaches of Mann and Candocia in the aspect of implementation flexibility. The proposed algorithm performs a better registration for real images than the case of registration that does not consider the exposure difference.

• Smart Structures and Systems
• /
• v.4 no.6
• /
• pp.809-828
• /
• 2008

Real-time hybrid testing is an attractive method to evaluate the response of structures under earthquake loads. The method is a variation of the pseudodynamic testing technique in which the experiment is executed in real time, thus allowing investigation of structural systems with time-dependent components. Real-time hybrid testing is challenging because it requires performance of all calculations, application of displacements, and acquisition of measured forces, within a very small increment of time. Furthermore, unless appropriate compensation for time delays and actuator time lag is implemented, stability problems are likely to occur during the experiment. This paper presents an approach for real-time hybrid testing in which time delay/lag compensation is implemented using model-based response prediction. The efficacy of the proposed strategy is verified by conducting substructure real-time hybrid testing of a steel frame under earthquake loads. For the initial set of experiments, a specimen with linear-elastic behavior is used. Experimental results agree well with the analytical solution and show that the proposed approach and testing system are capable of achieving a time-scale expansion factor of one (i.e., real time). Additionally, the proposed method allows accurate testing of structures with larger frequencies than when using conventional time delay compensation methods, thus extending the capabilities of the real-time hybrid testing technique. The method is then used to test a structure with a rate-dependent energy dissipation device, a magnetorheological damper. Results show good agreement with the predicted responses, demonstrating the effectiveness of the method to test rate-dependent components.

• The Journal of the Acoustical Society of Korea
• /
• v.38 no.1
• /
• pp.51-55
• /
• 2019

This paper proposes a speech enhancement method utilizing the feature compensation gain for robust speech recognition performances in noisy environments. In this paper we propose a speech enhancement method utilizing the feature compensation gain which is obtained from the PCGMM (Parallel Combined Gaussian Mixture Model)-based feature compensation method employing variational model composition. The experimental results show that the proposed method significantly outperforms the conventional front-end algorithms and our previous research over various background noise types and SNR (Signal to Noise Ratio) conditions in mismatched ASR (Automatic Speech Recognition) system condition. The computation complexity is significantly reduced by employing the noise model selection technique with maintaining the speech recognition performance at a similar level.

• Journal of Advanced Marine Engineering and Technology
• /
• v.18 no.4
• /
• pp.102-111
• /
• 1994

This article deals with the problem of designing a robust algorithm for the motion control of robot manipulator whose nonlinear dynamics contain various uncertainties. To ensure high performance of control system, a model-based feedforward compensation with continuous robust control has been developed. The control structure based on the deterministic approach consists of two parts : the nominal control law is first introduced to stabilize the system without uncertainties, then a robust nonlinear control law is adopted to compensate for both the resulting errors(or structured uncertainties) and unstructured uncertainties. The uncertainties assumed in this study are bounded by polynomials in the Euclidean norms of system states with known bounding coefficients. The presented control scheme is relatively simple as well as computationally efficient. With a feasible class of desired trajectories, the proposed control law provides sufficient criteria which guarantee that all possible responses of the closed-loop system are uniformly ultimately bounded in the presence of uncertainties. Therefore, the control algorithm proposed is shown to be robust with respect to the involved uncertainties.

• Proceedings of the Korean Society for Noise and Vibration Engineering Conference
• /
• 2005.11a
• /
• pp.213-216
• /
• 2005

This paper presents force feedback control performance of a 3DOF haptic device that can be used for minimally invasive surgery (MIS). As a first step, a 3DOF electrorheological (ER) joint is designed using a spherical mechanism. And it is optimized based on the mathematical torque modeling. Subsequently, the master haptic device is manufactured by the spherical joint. In order to achieve desired force trajectories, model based compensation strategy is adopted for the ER master. Therefore, Preisach model fur the PMA-based ER fluid is identified using experimental first order descending (FOD) curves. A compensation strategy is then formulated through the model inversion to achieve desired force at the ER master. Tracking control performances for sinusoidal force trajectory are presented, and their tracking errors are evaluated.

• Journal of Power Electronics
• /
• v.17 no.4
• /
• pp.983-990
• /
• 2017

This paper proposes a model predictive control based on the discrete Lyapunov function to improve the performance of power electronic converters. The proposed control technique, based on the finite control set model predictive control (FCS-MPC), defines a cost function for the control law which is determined under the Lyapunov stability theorem with a control error compensation. The steady state and dynamic performance of the proposed control strategy has been tested under a single phase AC/DC voltage source rectifier (S-VSR). Experimental results demonstrate that the proposed control strategy not only offers global stability and good robustness but also leads to a high quality sinusoidal current with a reasonably low total harmonic distortion (THD) and a fast dynamic response under linear loads.

• 전자공학회논문지 IE
• /
• v.44 no.3
• /
• pp.35-41
• /
• 2007

In this paper we suggest a model that the virtual robot based on artificial intelligence performs learning with compensation signals and compare the leaning speed of the virtual robot according to the compensation method after applying it to three type environments. As a result our model has showed that positive compensation is superior to hybrid one mixed positive and negative if there are enough time for learning in case of more or less complicated environment with the numerous foods, obstacles and robots. Otherwise hybrid method is better than positive one.

• Journal of the Korean Society of Manufacturing Technology Engineers
• /
• v.9 no.2
• /
• pp.102-111
• /
• 2000

Methodology of volumetric error identification and compensation is presented to improve the accuracy of NC machine tools by using a reference artifact and a touch trigger probe. Homogeneous transformation matrix and kinematic chain are used for modeling the geo-metric and thermal errors of a three-axis vertical machining center. The reference artifact is designed and fabricated to identify the model parameters by machine tool metrology. Parameters in the error model are able to be identified and updated by direct measurement of the reference artifact on the machine tool under the actual conditions which include the thermal interactions of error sources. A volumetric error compensation system based on IBM/PC is linked with a FANUC CNC controller to compensate for the identified volumetric error in machining workspace.

• Journal of the Korean Society for Precision Engineering
• /
• v.17 no.4
• /
• pp.135-141
• /
• 2000

The major role of temperature sensors in thermal error compensation system of machine tools is improving machining accuracy by supplying reliable temperature data on the machine structure. This paper presents a new method for fault diagnosis of temperature sensors and recovery of faulted data to establish the reliability of thermal error compensation system. The detection of fault and its location is based on the correlation coefficients among temperature data from the sensors. The multiple linear regression model which is prepared using complete normal data is also used fur the recovery of faulted data. The effectiveness of this method was tested by comparing the computer simulation results and measured data in a CNC machining center.