• ETRI Journal
• /
• v.32 no.5
• /
• pp.795-800
• /
• 2010

Embedding a large vocabulary speech recognition system in mobile devices requires a reduced acoustic model obtained by eliminating redundant model parameters. In conventional optimization methods based on the minimum description length (MDL) criterion, a binary Gaussian tree is built at each state of a hidden Markov model by iteratively finding and merging similar mixture components. An optimal subset of the tree nodes is then selected to generate a downsized acoustic model. To obtain a better binary Gaussian tree by improving the process of finding the most similar Gaussian components, this paper proposes a new distance measure that exploits the difference in likelihood values for cases before and after two components are combined. The mixture weight of Gaussian components is also introduced in the component merging step. Experimental results show that the proposed method outperforms MDL-based optimization using either a Kullback-Leibler (KL) divergence or weighted KL divergence measure. The proposed method could also reduce the acoustic model size by 50% with less than a 1.5% increase in error rate compared to a baseline system.

• Journal of Communications and Networks
• /
• v.16 no.3
• /
• pp.264-269
• /
• 2014

We consider transmission over body area networks. Due to the difficulty in assessing an accurate statistical model valid for multiple scenarios, we advocate a system design technique favoring robustness. Our approach, which is based on results in [12] and generalizes them, examines the variation of a performance metric when the nominal statistical distribution of fading is replaced by the worst distribution within a given Kullback-Leibler divergence from it. The sensitivity of the performance metric to the divergence from the nominal distribution can be used as an indication of the design robustness. This concept is applied by evaluating the error probability of binary uncoded modulation and the outage probability-the first parameter is useful to assess system performance with no error-control coding, while the second reflects the performance when a near-optimal code is used. The usefulness of channel coding can be assessed by comparing its robustness with that of uncoded transmission.

• Journal of the Korean Statistical Society
• /
• v.33 no.1
• /
• pp.129-147
• /
• 2004

For square contingency tables with nominal categories, this paper proposes a measure to represent the degree of departure from the quasi-symmetry (QS) model and the Bradley-Terry (BT) model. The measure proposed is expressed by using the Cressie and Read (1984)'s power-divergence or Patil and Taillie (1982)'s diversity index. The measure lies between 0 and 1, and it is useful for comparing the degree of departure from QS or BT in several tables.

• Journal of the Korean Statistical Society
• /
• v.27 no.3
• /
• pp.289-303
• /
• 1998

For square contingency tables with ordered categories, Tomizawa (1995) considered two kinds of measures to represent the degree of departure from global symmetry, which means that the probability that an observation will fall in one of cells in the upper-right triangle of square table is equal to the probability that the observation falls in one of cells in the lower-left triangle of it. This paper proposes a generalization of those measures. The proposed measure is expressed by using Cressie and Read's (1984) power divergence or Patil and Taillie's (1982) diversity index. Special cases of the proposed measure include TomiBawa's measures. The proposed measure would be useful for comparing the degree of departure from global symmetry in several tables.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.41 no.8
• /
• pp.851-858
• /
• 2016

This paper presents a new statistical voice activity detection (VAD) based on the probabilistic interpretation of nonnegative matrix factorization (NMF). The objective function of the NMF using Kullback-Leibler divergence coincides with the negative log likelihood function of the data if the distribution of the data given the basis and encoding matrices is modeled as Poisson distributions. Based on this probabilistic NMF, the VAD is constructed using the likelihood ratio test assuming that speech and noise follow Poisson distributions. Experimental results show that the proposed approach outperformed the conventional Gaussian model-based and NMF-based methods at 0-15 dB signal-to-noise ratio simulation conditions.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.20 no.4
• /
• pp.833-839
• /
• 2016

In this paper, we propose a selective feature extraction algorithm between Markov transition probability and co-occurrence probability for an effective image splicing detection. The Features used in our method are composed of the difference values between DCT coefficients in the adjacent blocks and the value of Kullback-Leibler divergence(KLD) is calculated to evaluate the differences between the distribution of original image features and spliced image features. KLD value is an efficient measure for selecting Markov feature or Co-occurrence feature because KLD shows non-similarity of the two distributions. After training the extracted feature vectors using the SVM classifier, we determine whether the presence of the image splicing forgery. To verify our algorithm we used grid search and 6-folds cross-validation. Based on the experimental results it shows that the proposed method has good detection performance with a limited number of features compared to conventional methods.

• Journal of KIISE:Computing Practices and Letters
• /
• v.15 no.6
• /
• pp.405-413
• /
• 2009

Tagging system is the system that allows internet users to assign new meta-data which is called tag to article, photo, video and etc. for facilitating searching and browsing of web contents. Tag cloud, a visual interface is widely used for browsing tag space. Tag cloud selects the tags with the highest frequency and presents them alphabetically with font size reflecting their popularity. However the conventional tag selection method includes known weaknesses. So, we propose a novel tag selection method Freshness, which helps to find fresh web contents. Freshness is the mean value of Kullback-Leibler divergences between each consecutive change of tag co-occurrence probability distribution. We collected tag data from three web sites, Allblog, Eolin and Technorati and constructed the system, 'Fresh Tag Cloud' which collects tag data and creates our tag cloud. Comparing the experimental results between Fresh Tag Cloud and the conventional one with data from Allblog, our one shows 87.5% less overlapping average, which means Fresh Tag Cloud outperforms the conventional tag cloud.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2017.06a
• /
• pp.250-253
• /
• 2017

본 논문에서는 동시적 위치 추정 및 지도 작성 (simultaneous localization and mapping)에서 루프 폐쇄 검출을 딥러닝 방법의 일종인 variational autoencoder 를 이용하여 수행하는 방법에 대해 살펴본다. Autoencoder 는 비감독 학습 방법의 일종으로 입력 영상이 신경망을 통과하여 얻은 출력 영상과 동일하도록 신경망을 학습시키는 모델이다. 이 때 autoencoder 중간의 병목 지역을 통과함에도 불구하고 입력과 동일한 영상을 계산해야 하는 제약조건이 있기 때문에 이는 차원 축소나 데이터 추상화의 목적으로 많이 사용된다. 여기서 한 단계 더 발전된 variational autoencoder 는 기존의 autoencoder 가 가진 단점인 입력 변수의 분포와 잠재 변수의 분포 사이에 상관관계가 없다는 단점을 해결하기 위해 Kullback-Leibler divergence 를 활용한 손실 함수를 정의하여 사용했다. 실험결과에서는 루프 폐쇄 검출에서 많이 사용되는 City-Centre 와 New College 데이터 집합을 사용하여 평가하였으며 루프 폐쇄 검출의 결과는 정밀도와 재현율을 계산하여 나타냈다.

• Proceedings of the Korean Operations and Management Science Society Conference
• /
• 2003.05a
• /
• pp.729-733
• /
• 2003

The presence of local features within clusters incurred by multi-modal nature of data prohibits many conventional clustering techniques from working properly. Especially, the clustering of datasets with non-Gaussian distributions within a cluster can be problematic when the technique with implicit assumption of Gaussian distribution is used. Current study proposes a simple tandem clustering method composed of k-means type algorithm and hierarchical method to solve such problems. The multi-modal dataset is first divided into many small pre-clusters by k-means or fuzzy k-means algorithm. The pre-clusters found from the first step are to be clustered again using agglomerative hierarchical clustering method with Kullback- Leibler divergence as the measure of dissimilarity. This method is not only effective at extracting the multi-modal clusters but also fast and easy in terms of computation complexity and relatively robust at the presence of outliers. The performance of the proposed method was evaluated on three generated datasets and six sets of publicly known real world data.

• Journal of Korean Institute of Industrial Engineers
• /
• v.42 no.2
• /
• pp.86-95
• /
• 2016

We propose a new approach to the stochastic version of Lanchester model. Commonly used approach to stochastic Lanchester model is through the Markov-chain method. The Markov-chain approach, however, is not appropriate to high dimensional heterogeneous force case because of large computational cost. In this paper, we propose an approximation method of stochastic Lanchester model. By matching the first and the second moments, the distribution of each unit strength can be approximated with multivariate normal distribution. We evaluate an approximation of discrete Markov-chain model by measuring Kullback-Leibler divergence. We confirmed high accuracy of approximation method, and also the accuracy and low computational cost are maintained under high dimensional heterogeneous force case.