• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.45 no.5
• /
• pp.1-8
• /
• 2008

The belief propagation method that has been studied recently yields good performance in disparity extraction. The method in which a target function is modeled as an energy function based on Markov random field(MRF), solves the stereo matching problem by finding the disparity to minimize the energy function. MRF models provide robust and unified framework for vision problem such as stereo and image restoration. the belief propagation method produces quite correct results, but it has difficulty in real time implementation because of higher computational complexity than other stereo methods. To relieve this problem, in this paper, we propose a fast algorithm of the belief propagation method. Energy function consists of a data term and a smoothness tern. The data term usually corresponds to the difference in brightness between correspondences, and smoothness term indicates the continuity of adjacent pixels. Smoothness information is created from messages, which are assigned using four different message arrays for the pixel positions adjacent in four directions. The processing time for four message arrays dominates 80 percent of the whole program execution time. In the proposed method, we propose an algorithm that dramatically reduces the processing time require in message calculation, since the message.; are not produced in four arrays but in a single array. Tn the last step of disparity extraction process, the messages are called in the single integrated array and this algorithm requires 1/4 computational complexity of the conventional method. Our method is evaluated by comparing the disparity error rates of our method and the conventional method. Experimental results show that the proposed method remarkably reduces the execution time while it rarely increases disparity error.

• Phonetics and Speech Sciences
• /
• v.12 no.3
• /
• pp.65-71
• /
• 2020

Feature normalization is a method to reduce the effect of environmental mismatch between the training and test conditions through the normalization of statistical characteristics of acoustic feature parameters. It demonstrates excellent performance improvement in the traditional Gaussian mixture model-hidden Markov model (GMM-HMM)-based speech recognition system. However, in a deep neural network (DNN)-based speech recognition system, minimizing the effects of environmental mismatch does not necessarily lead to the best performance improvement. In this paper, we attribute the cause of this phenomenon to information loss due to excessive feature normalization. We investigate whether there is a feature normalization method that maximizes the speech recognition performance by properly reducing the impact of environmental mismatch, while preserving useful information for training acoustic models. To this end, we introduce the mean and exponentiated variance normalization (MEVN), which is a compromise between the mean normalization (MN) and the mean and variance normalization (MVN), and compare the performance of DNN-based speech recognition system in noisy and reverberant environments according to the degree of variance normalization. Experimental results reveal that a slight performance improvement is obtained with the MEVN over the MN and the MVN, depending on the degree of variance normalization.

• KDI Journal of Economic Policy
• /
• v.36 no.2
• /
• pp.103-136
• /
• 2014

The covered interest rate parity condition (CIRP) has been widely used in open macroeconomic analysis, risk management, exchange rate forecasts, and so forth. Due to the recent global financial crises, there have been remarkable changes in the financial markets of the emerging markets. These changes possibly influenced the dynamics of the covered interest rate parity condition. In this paper, we investigate whether the CIRP dynamics has changed, and what is the nature of the regime changes. To do this, we propose and estimate multiple-state Markov regime switching models using a Bayesian MCMC method. Our estimation results indicate that the default risk or the deviation from the CIRP has been decreased after the crisis. It seems to be associated with the more active interaction between the short-term bond market and the short-term foreign exchange market than before. The tightened relation of these two financial markets is caused by the arbitrage transaction of foreign investors.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.26 no.6
• /
• pp.471-476
• /
• 2016

In this paper, recognition system for continuous human action is explained by using motion history image and histogram of oriented gradient with spotter model based on depth information, and the spotter model which performs action spotting is proposed to improve recognition performance in the recognition system. The steps of this system are composed of pre-processing, human action and spotter modeling and continuous human action recognition. In pre-processing process, Depth-MHI-HOG is used to extract space-time template-based features after image segmentation, and human action and spotter modeling generates sequence by using the extracted feature. Human action models which are appropriate for each of defined action and a proposed spotter model are created by using these generated sequences and the hidden markov model. Continuous human action recognition performs action spotting to segment meaningful action and meaningless action by the spotter model in continuous action sequence, and continuously recognizes human action comparing probability values of model for meaningful action sequence. Experimental results demonstrate that the proposed model efficiently improves recognition performance in continuous action recognition system.

• KIPS Transactions on Software and Data Engineering
• /
• v.11 no.8
• /
• pp.331-338
• /
• 2022

Successor representation (SR) is a model of human reinforcement learning (RL) mimicking the underlying mechanism of hippocampal cells constructing cognitive maps. SR utilizes these learned features to adaptively respond to the frequent reward changes. In this paper, we evaluated the performance of SR under the context where changes in latent variables of environments trigger the reward structure changes. For a benchmark test, we adopted SR-Dyna, an integration of SR into goal-driven Dyna RL algorithm in the 2-stage Markov Decision Task (MDT) in which we can intentionally manipulate the latent variables - state transition uncertainty and goal-condition. To precisely investigate the characteristics of SR, we conducted the experiments while controlling each latent variable that affects the changes in reward structure. Evaluation results showed that SR-Dyna could learn to respond to the reward changes in relation to the changes in latent variables, but could not learn rapidly in that situation. This brings about the necessity to build more robust RL models that can rapidly learn to respond to the frequent changes in the environment in which latent variables and reward structure change at the same time.

• The Journal of the Acoustical Society of Korea
• /
• v.40 no.5
• /
• pp.530-536
• /
• 2021

Recently, due to the development of deep learning, end-to-end speech recognition, which directly maps graphemes to speech signals, shows good performance. Especially, among the end-to-end models, conformer shows the best performance. However end-to-end models only focuses on the probability of which grapheme will appear at the time. The decoding process uses a greedy search or beam search. This decoding method is easily affected by the final probability output by the model. In addition, the end-to-end models cannot use external pronunciation and language information due to structual problem. Therefore, in this paper conformer with lexicon transducer is proposed. We compare phoneme-based model with lexicon transducer and grapheme-based model with beam search. Test set is consist of words that do not appear in training data. The grapheme-based conformer with beam search shows 3.8 % of CER. The phoneme-based conformer with lexicon transducer shows 3.4 % of CER.

• Management & Information Systems Review
• /
• v.34 no.3
• /
• pp.17-39
• /
• 2015

This study examined the structural changes and volatility in the global stock markets using a Markov Regime Switching ARCH model developed by the Hamilton and Susmel (1994). Firstly, the US, Italy and Ireland showed that variance in the high volatility regime was more than five times that in the low volatility, while Korea, Russia, India, and Greece exhibited that variance in the high volatility regime was increased more than eight times that in the low. On average, a jump from regime 1 to regime 2 implied roughly three times increased in risk, while the risk during regime 3 was up to almost thirteen times than during regime 1 over the study period. And Korea, the US, India, Italy showed ARCH(1) and ARCH(2) effects, leverage and asymmetric effects. Secondly, 278 days were estimated in the persistence of low volatility regime, indicating that the mean transition probability between volatilities exhibited the highest long-term persistence in Korea. Thirdly, the coefficients appeared to be unstable structural changes and volatility for the stock markets in Chow tests during the Asian, Global and European financial crisis. In addition, 1-Step prediction error tests showed that stock markets were unstable during the Asian crisis of 1997-1998 except for Russia, and the Global crisis of 2007-2008 except for Korea and the European crisis of 2010-2011 except for Korea, the US, Russia and India. N-Step tests exhibited that most of stock markets were unstable during the Asian and Global crisis. There was little change in the Asian crisis in CUSUM tests, while stock markets were stable until the late 2000s except for some countries. Also there were stable and unstable stock markets mixed across countries in CUSUMSQ test during the crises. Fourthly, I confirmed a close relevance of the volatility between Korea and other countries in the stock markets through the likelihood ratio tests. Accordingly, I have identified the episode or events that generated the high volatility in the stock markets for the financial crisis, and for all seven stock markets the significant switch between the volatility regimes implied a considerable change in the market risk. It appeared that the high stock market volatility was related with business recession at the beginning in 1990s. By closely examining the history of political and economical events in the global countries, I found that the results of Lamoureux and Lastrapes (1990) were consistent with those of this paper, indicating there were the structural changes and volatility during the crises and specificly every high volatility regime in SWARCH-L(3,2) student t-model was accompanied by some important policy changes or financial crises in countries or other critical events in the international economy. The sophisticated nonlinear models are needed to further analysis.

• KSCE Journal of Civil and Environmental Engineering Research
• /
• v.43 no.4
• /
• pp.421-431
• /
• 2023

Sewer deterioration models can offer important information on prediction of future condition of the asset to decision makers in their implementing sewer pipe networks management program. In this study, Markov chain model was used to estimate sewer deterioration trend based on the historical structural condition assessment data obtained by CCTV inspection. The data used in this study were limited to Hume pipe with diameter of 450 mm and 600 mm in three sub-catchment areas in city A, which were collected by CCTV inspection projects performed in 1998-1999 and 2010-2011. As a result, it was found that sewers in sub-catchment area EM have deteriorated faster than those in other two sub-catchments. Various main defects were to generate in 29% of 450 mm sewers and 38% of 600 mm in 35 years after the installation, while serious failure in 62% of 450 mm sewers and 74% of 600 mm in 100 years after the installation in sub-catchment area EM. In sub-catchment area SN, main defects were to generate in 26% of 450 mm sewers and 35% of 600 mm in 35 years after the installation, while in sub-catchment area HK main defects were to generate in 27% of 450 mm sewers and 37% of 600 mm in 35 years after the installation. Larger sewer pipes of 600 mm were found to deteriorate faster than smaller sewer pipes of 450 mm by about 12 years. Assuming that the percentage of main defects generation could be set as 40% to estimate the life expectancy of the sewers, it was estimated as 60 years in sub-catchment area SN, 42 years in sub-catchment area EM, 59 years in sub-catchment area HK for 450 mm sewer pipes, respectively. For 600 mm sewer pipes, on the other hand, it was estimated as 43 years, 34 years, 39 years in sub-catchment areas SN, EM, and HK, respectively.

• The Journal of the Acoustical Society of Korea
• /
• v.21 no.2
• /
• pp.199-210
• /
• 2002

In this paper, we carried out the study on speech recognition using the KM-Net topology design algorithm based on decision tree state-clustering to improve the performance of acoustic models in speech recognition. The Korean has many allophonic and grammatical rules compared to other languages, so we investigate the allophonic variations, which defined the Korean phonetics, and construct the phoneme question set for phonetic decision tree. The basic idea of the HM-Net topology design algorithm is that it has the basic structure of SSS (Successive State Splitting) algorithm and split again the states of the context-dependent acoustic models pre-constructed. That is, it have generated. the phonetic decision tree using the phoneme question sets each the state of models, and have iteratively trained the state sequence of the context-dependent acoustic models using the PDT-SSS (Phonetic Decision Tree-based SSS) algorithm. To verify the effectiveness of the above algorithm we carried out the speech recognition experiments for 452 words of center for Korean language Engineering (KLE452) and 200 sentences of air flight reservation task (YNU200). Experimental results show that the recognition accuracy has progressively improved according to the number of states variations after perform the splitting of states in the phoneme, word and continuous speech recognition experiments respectively. Through the experiments, we have got the average 71.5%, 99.2% of the phoneme, word recognition accuracy when the state number is 2,000, respectively and the average 91.6% of the continuous speech recognition accuracy when the state number is 800. Also we haute carried out the word recognition experiments using the HTK (HMM Too1kit) which is performed the state tying, compared to share the parameters of the HM-Net topology design algorithm. In word recognition experiments, the HM-Net topology design algorithm has an average of 4.0% higher recognition accuracy than the context-dependent acoustic models generated by the HTK implying the effectiveness of it.

• The Journal of the Acoustical Society of Korea
• /
• v.30 no.3
• /
• pp.149-157
• /
• 2011

In this paper, we propose a new cardiac disorder classification method using an support vector machine (SVM) to combine hidden Markov model (HMM) and murmur existence information. Using cepstral features and the HMM Viterbi algorithm, we segment input heart sound signals into HMM states for each cardiac disorder model and compute log-likelihood (score) for every state in the model. To exploit the temporal position characteristics of murmur signals, we divide the input signals into two subbands and compute murmur probability of every subband of each frame, and obtain the murmur score for each state by using the state segmentation information obtained from the Viterbi algorithm. With an input vector containing the HMM state scores and the murmur scores for all cardiac disorder models, SVM finally decides the cardiac disorder category. In cardiac disorder classification experimental results, the proposed method shows the relatively improvement rate of 20.4 % compared to the HMM-based classifier with the conventional cepstral features.