• Proceedings of the KSPS conference
• /
• 2005.04a
• /
• pp.87-90
• /
• 2005

Today's state-of the-art speech recognition systems typically use continuous distribution hidden Markov models with the mixtures of Gaussian distributions. To obtain higher recognition accuracy, the hidden Markov models typically require huge number of Gaussian distributions. Such speech recognition systems have problems that they require too much memory to run, and are too slow for large applications. Many approaches are proposed for the design of compact acoustic models. One of those models is subspace distribution clustering hidden Markov model. Subspace distribution clustering hidden Markov model can represent original full-space distributions as some combinations of a small number of subspace distribution codebooks. Therefore, how to make the codebook is an important issue in this approach. In this paper, we report some experimental results on various quantization methods to make more accurate models.

• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.32B no.12
• /
• pp.1663-1669
• /
• 1995

This paper is a study on the recognition of Korean proverb using neural network and Markov model. The neural network uses, at the stage of training neurons, features such as the rate of zero crossing, short-term energy and PLP-Cepstrum, covering a time of 300ms long. Markov models were generated by the recognized phoneme strings. The recognition of words and proverbs using Markov models have been carried out. Experimental results show that phoneme and word recognition rates are 81. 2%, 94.0% respectively for Korean proverb recognition experiments.

• Korean Management Science Review
• /
• v.10 no.2
• /
• pp.27-42
• /
• 1993

Markov chain models can be used to predict the state of the system in the future. We extend the existing Markov chain models in two ways. For the stationary model, we propose a procedure that obtains the transition probabilities by appling the empirical Bayes method, in which the parameters of the prior distribution in the Bayes estimator are obtained on the collaternal micro data. For non-stationary model, we suggest a procedure that obtains a time-varying transition probabilities as a function of the exogenous variables. To illustrate the effectiveness of our extended models, the models are applied to the macro and micro time-series data generated from actual survey. Our stationary model yields reliable parameter values of the prior distribution. And our non-stationary model can predict the variable transition probabilities effectively.

• The Journal of the Acoustical Society of Korea
• /
• v.8 no.1
• /
• pp.81-87
• /
• 1989

This paper discribes the method associated with modeling of Korean phonemes. Hidden Markov models(HMM's) may be viewed as an effective technique for modeling the inherent nonstationarity of speech signal. We propose a 3-state phoneme model to represent the sequentially changing characteristics of phonemes, i.e., transition-to-stationary-to-transition. Also we clarify that the duration of a phoneme is an important factor to have an effect in recognition accuracy and show that improvement in recognition rate can be obtained by using duration-dependent 3-state hidden Markov models.

• Journal of the Korean Institute of Telematics and Electronics S
• /
• v.34S no.2
• /
• pp.71-78
• /
• 1997

Markov random field models have extensively used in applications such as image segmentation and image restoration. In this paper, we consider the relation between the stationarity of parameters and the synthesized images for gauss-markov rnadom field which has the most popularly used among many MRF models. GMRF model, which is both wide-sense Markov and strict-sense markov, has AR representations and is also a kind of gibbs distribution. Therefore, we may approach in aspect of both AR models and gibbs models. We show the relation between the stationarity of parameters and the images which are synthesized by two approaching methods and derive the stationary regions of parameters in 1st order and isotropic 2nd order case.

• International Journal of Computer Science & Network Security
• /
• v.21 no.8
• /
• pp.28-34
• /
• 2021

Internet Service Providers (ISPs) constantly endeavor to resolve network congestion, in order to provide fast and cheap services to the customers. This study suggests two models based on Markov chain, using three and four access attempts to complete the call. It involves a comparative study of four models to check the relationship between Internet Access sharing traffic, and the possibility of network jamming. The first model is a Markov chain, based on call-by-call attempt, whereas the second is based on two attempts. Models III&IV suggested by the authors are based on the assumption of three and four attempts. The assessment reveals that sometimes by increasing the number of attempts for the same operator, the chances for the customers to complete the call, is also increased due to blocking probabilities. Three and four attempts express the actual relationship between traffic sharing and blocking probability based on Markov using MATLAB tools with initial probability values. The study reflects shouting results compared to I&II models using one and two attempts. The success ratio of the first model is 84.5%, and that of the second is 90.6% to complete the call, whereas models using three and four attempts have 94.95% and 95.12% respectively to complete the call.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2015.05a
• /
• pp.488-488
• /
• 2015

Intermittent streamflow is common phenomenon in arid and semi-arid regions. To manage water resources of intermittent streamflows, stochactic simulation data is essential; however the seasonally stochastic modeling for intermittent streamflow is a difficult task. In this study, using the periodic Markov chain model, we simulate intermittent monthly streamflow for occurrence and the periodic gamma autoregressive and copula models for amount. The copula models were tested in a previous study for the simulation of yearly streamflow, resulting in successful replication of the key and operational statistics of historical data; however, the copula models have never been tested on a monthly time scale. The intermittent models were applied to the Colorado River system in the present study. A few drawbacks of the PGAR model were identified, such as significant underestimation of minimum values on an aggregated yearly time scale and restrictions of the parameter boundaries. Conversely, the copula models do not present such drawbacks but show feasible reproduction of key and operational statistics. We concluded that the periodic Markov chain based the copula models is a practicable method to simulate intermittent monthly streamflow time series.

• Journal of Institute of Control, Robotics and Systems
• /
• v.18 no.11
• /
• pp.1005-1010
• /
• 2012

We propose a novel approach to classify fingerprints using the extracted adaptive core block for improving classification performance of incomplete fingerprints in this paper. We compute representative directions from fingerprint images by the block unit and learn horizontal and vertical Markov models by deciding the center position of a fingerprint image based on the expert knowledge. The center block of a test image is the block has the highest probability after comparing the Markov model with $11{\times}11$ blocks. The proposed approach can effectively classify incomplete fingerprints using the optimal center block.

• Journal of Applied Reliability
• /
• v.16 no.4
• /
• pp.313-322
• /
• 2016

Purpose: We introduce ways to employ Markov chain model to evaluate the effect of preventive maintenance process. While the preventive maintenance process decreases the failure rate of each subsystems, it increases the downtime of the system because the system can not work during the maintenance process. The goal of this paper is to introduce ways to analyze this trade-off. Methods: Markov chain models are employed. We derive the availability of the system consisting of N repairable subsystems by the methods under various maintenance policies. Results: To validate our methods, we apply our models to the real maintenance data reports of military truck. The error between the model and the data was about 1%. Conclusion: The models developed in this paper fit real data well. These techniques can be applied to calculate the availability under various preventive maintenance policies.

• Journal of Korea Multimedia Society
• /
• v.20 no.9
• /
• pp.1502-1508
• /
• 2017

A musical key is a fundamental concept in Western music theory. It is a collective characterization of pitches and chords that together create a musical perception of the entire piece. It is based on a group of pitches in a scale with which a music is constructed. Each key specifies the set of seven primary chromatic notes that are used out of the twelve possible notes. This paper presents a method that identifies the key of a song using Hidden Markov Models given a sequence of chroma features. Given an input song, a sequence of chroma features are computed. It is then classified into one of the 24 keys using a discrete Hidden Markov Models. The proposed method can help musicians and disc-jockeys in mixing a segment of tracks to create a medley. When tested on 120 songs, the success rate of the music key identification reached around 87.5%.