• Proceedings of the Acoustical Society of Korea Conference
• /
• spring
• /
• pp.337-338
• /
• 2000

The perceptual filter for speech enhancement was analytically derived where the frequency content of the input noisy signal was made the same as that of the estimated clean signal in auditory domain. However, the analytical derivation should rely on the deconvolution associated with the spreading function in the psychoacoustic model, which results in an ill-conditioned problem. In order to cope with the problem associated with the deconvolution, we propose a novel psychoacoustic model based speech enhancement filter whose principle is the same as the perceptual filter, however the filter is derived by a constrained optimization which provides solutions to the ill-conditioned problem.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• v.13 no.3
• /
• pp.164-170
• /
• 2013

Recent years have been higher demands for automatic speech recognition (ASR) systems that are able to operate robustly in an acoustically noisy environment. This paper proposes an improved product hidden markov model (HMM) used for bimodal speech recognition. A two-dimensional training model is built based on dependently trained audio-HMM and visual-HMM, reflecting the asynchronous characteristics of the audio and video streams. A weight coefficient is introduced to adjust the weight of the video and audio streams automatically according to differences in the noise environment. Experimental results show that compared with other bimodal speech recognition approaches, this approach obtains better speech recognition performance.

• ETRI Journal
• /
• v.33 no.6
• /
• pp.914-923
• /
• 2011

This paper concerns robust and reliable speaker model training for text-independent speaker verification. The baseline speaker modeling approach is the Gaussian mixture model (GMM). In text-independent speaker verification, the amount of speech data may be different for speakers. However, we still wish the modeling approach to perform equally well for all speakers. Besides, the modeling technique must be least vulnerable against unseen data. A traditional approach for GMM training is expectation maximization (EM) method, which is known for its overfitting problem and its weakness in handling insufficient training data. To tackle these problems, variational approximation is proposed. Variational approaches are known to be robust against overtraining and data insufficiency. We evaluated the proposed approach on two different databases, namely KING and TFarsdat. The experiments show that the proposed approach improves the performance on TFarsdat and KING databases by 0.56% and 4.81%, respectively. Also, the experiments show that the variationally optimized GMM is more robust against noise and the verification error rate in noisy environments for TFarsdat dataset decreases by 1.52%.

• The Journal of the Acoustical Society of Korea
• /
• v.19 no.3E
• /
• pp.12-18
• /
• 2000

Psychoacoustic model based methods have recently been introduced in order to enhance speech signals corrupted by ambient noise. In particular, the perceptual filter is analytically derived where the frequency content of the input noisy signal is made the same as that of the estimated clean signal in auditory domain. However, the analytical derivation should rely on the deconvolution associated with the spreading function in the psychoacoustic model, which results in an ill-conditioned problem. In order to cope with the problem associated with the deconvolution, we propose a novel psychoacoustic model based speech enhancement filter whose principle is the same as the perceptual filter, however the filter is derived by a constrained optimization which provides solutions to the ill-conditioned problem. It is demonstrated with artificially generated signals that the proposed filter operates according to the principle. It is shown that superior performance results from the proposed filter over the perceptual filter provided that a clean speech signal is separable from noise.

• IEMEK Journal of Embedded Systems and Applications
• /
• v.18 no.2
• /
• pp.81-87
• /
• 2023

With the development of BCI devices, it is now possible to use EEG control technology to move the robot's arms or legs to help with daily life. In this paper, we propose a customized vehicle control model based on BCI. This is a model that collects BCI-based driver EEG signals, determines information according to EEG signal analysis, and then controls the direction of the vehicle based on the determinated information through EEG signal analysis. In this case, in the process of analyzing noisy EEG signals, controlling direction is supplemented by using a camera-based eye tracking method to increase the accuracy of recognized direction . By synthesizing the EEG signal that recognized the direction to be controlled and the result of eye tracking, the vehicle was controlled in five directions: left turn, right turn, forward, backward, and stop. In experimental result, the accuracy of direction recognition of our proposed model is about 75% or higher.

• ETRI Journal
• /
• v.26 no.3
• /
• pp.273-276
• /
• 2004

We propose a novel feature processing technique which can provide a cepstral liftering effect in the log-spectral domain. Cepstral liftering aims at the equalization of variance of cepstral coefficients for the distance-based speech recognizer, and as a result, provides the robustness for additive noise and speaker variability. However, in the popular hidden Markov model based framework, cepstral liftering has no effect in recognition performance. We derive a filtering method in log-spectral domain corresponding to the cepstral liftering. The proposed method performs a high-pass filtering based on the decorrelation of filter-bank energies. We show that in noisy speech recognition, the proposed method reduces the error rate by 52.7% to conventional feature.

• Proceedings of the IEEK Conference
• /
• 2001.06d
• /
• pp.29-32
• /
• 2001

Visual speech information improves the performance of speech recognition, especially in noisy environment. We have tested the various spatial-temporal features for the Korean lipreading and evaluated the performance by using a hidden Markov model based classifier. The results have shown that the direction as well as the magnitude of the movement of the lip contour over time is useful features for the lipreading.

• Journal of applied mathematics & informatics
• /
• v.8 no.2
• /
• pp.613-625
• /
• 2001

In this paper, we investigate a simplified model of a particle suspended elastically from two towers by two nonlinear elastic springs, with a restoring force similar to Hooke’s law under extension and with no resistance to compression. Numerical results are presented, showing the solutions can be either of the same period oscillation the forcing term, can be a subharmonic response of multiple period, or can be noisy periodic which is apparently chaotic. Multiplicity of periodic solutions for certain physical parameters are demonstrated.

• Journal of the Korean Operations Research and Management Science Society
• /
• v.3 no.2
• /
• pp.41-45
• /
• 1978

Duel models are frequently used in war game simulation. Both game-theoretic approach and stochastic approach are applied to duel situations in war game. Game-theoretic models are usually classified into three categories, noisy duel, silent duel, and duel of continuous firing. Stochastic duels are classified depending upon assumptions. In this paper formulation and a general solution for each model will be summarized.

• 제어로봇시스템학회:학술대회논문집
• /
• 1996.10b
• /
• pp.892-895
• /
• 1996

An unknown input observer is proposed that can be used in wheelbase preview control of commercial vehicles. The preview and state information, required to calculate actuator force, are reconstructed from the measurement variables such as heave and pitch acceleration. Gain matrix of observer is optimally selected so that influence of system and measurement noises on the estimation error can be minimized. Estimated preview information requires low pass filtering to eliminate high frequency components resulting from differentiation of noisy output signals. Effectiveness of the proposed method is demonstrated by numerical simulation of half car model.