• Phonetics and Speech Sciences
• /
• v.1 no.2
• /
• pp.3-8
• /
• 2009

In an effort to improve speech recognition, we investigated the effect of the number of clusters. In usual LBG clustering, the number of codebook clusters is doubled on each bifurcation and hence cannot be chosen arbitrarily in a natural way. To have the number of clusters at our control, we combined adaptive resonance theory (ART2) with LBG and perform the clustering in two stages. The codebook thus formed was used in subsequent processing of fuzzy vector quantization (FVQ) and HMM for speech recognition tests. Compared to conventional LBG, our method was shown to reduce the best recognition error rate by 0${\sim$}0.9% depending on the vocabulary size. The result also showed that between 400 and 800 would be the optimal number of clusters in the limit of small and large vocabulary speech recognitions of isolated words, respectively.

• Proceedings of the Korean Information Science Society Conference
• /
• 2006.10b
• /
• pp.233-237
• /
• 2006

This study proposes an innovative measure for evaluating the performance of text clustering. In using K-means algorithm and Kohonen Networks for text clustering, the number clusters is fixed initially by configuring it as their parameter, while in using single pass algorithm for text clustering, the number of clusters is not predictable. Using labeled documents, the result of text clustering using K-means algorithm or Kohonen Network is able to be evaluated by setting the number of clusters as the number of the given target categories, mapping each cluster to a target category, and using the evaluation measures of text. But in using single pass algorithm, if the number of clusters is different from the number of target categories, such measures are useless for evaluating the result of text clustering. This study proposes an evaluation measure of text clustering based on intra-cluster similarity and inter-cluster similarity, what is called CI (Clustering Index) in this article.

• Journal of the Korean Statistical Society
• /
• v.29 no.2
• /
• pp.247-260
• /
• 2000

This paper is concerned with a modification of the k-spatial medians clustering. To find a suitable number of clusters, the number k of clusters is incorporated into the k-spatial medians clustering criterion through a weight function. Proposed method for the choice of the weight function offers a reasonable number of clusters. Some theoretical properties of the method are investigated along with some examples.

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.41 no.6
• /
• pp.39-46
• /
• 2004

One of the significant issues in cluster analysis is to identify a proper number of clusters hidden under given data. In this paper we propose a novel approach to systematically determine the number of clusters based on Input Representation Coverage (IRC), which is newly defined as a quantified value of how well original input data in Gaussian feature space can be captured with a certain number of clusters. Furthermore, its usability and applicability is also investigated via experiments with synthetic data. Our experiment results show that the proposed approach is quite useful in approximately finding the real number of clusters implicitly contained in the data.

• International journal of advanced smart convergence
• /
• v.7 no.2
• /
• pp.33-37
• /
• 2018

In this study, we propose an iterative clustering algorithm that automatically clusters a set of voice signal data without a label into an optimal number of clusters and generates hmm model for each cluster. In the clustering process, the likelihood calculations of the clusters are performed using iterative hmm learning and testing while varying the number of clusters for given data, and the maximum likelihood estimation method is used to determine the optimal number of clusters. We tested the effectiveness of this clustering algorithm on a small-vocabulary digit clustering task by mapping the unsupervised decoded output of the optimal cluster to the ground-truth transcription, we found out that they were highly correlated.

• Journal of the Korean Operations Research and Management Science Society
• /
• v.31 no.2
• /
• pp.87-98
• /
• 2006

In many cases of cluster analysis we are forced to perform clustering without any prior knowledge on the number of clusters. But in some clustering methods such as k-means algorithm it is required to provide the number of clusters beforehand. In this study, we focus on the problem to determine the number of clusters in the given data. We follow the 2 stage approach of ASA clustering algorithm and mainly try to improve the performance of the first stage of the algorithm. We verify the usefulness of the method by applying it for various kinds of simulated data. Also, we apply the method for clustering two kinds of real life qualitative data.

• Journal of Digital Contents Society
• /
• v.12 no.3
• /
• pp.339-345
• /
• 2011

The k-means has been one of the popular, simple and faster clustering algorithms, but the right value of k is unknown. The value of k (the number of clusters) is a very important element because the result of clustering is different depending on it. In this paper, we present a novel algorithm based on an external cluster relationship criterion which is an evaluation metric of clustering result to determine the number of clusters dynamically. Experimental results show that our algorithm is superior to other methods in terms of the accuracy of the number of clusters.

• Journal of the Korean Operations Research and Management Science Society
• /
• v.42 no.3
• /
• pp.25-34
• /
• 2017

K-means is a popular and efficient data clustering method that only uses intra-cluster distance to establish a valid index with a previously fixed number of clusters. K-means is useless without a suitable number of clusters for unsupervised data. This paper aimsto propose the Group Search Optimization (GSO) using Silhouette to find the optimal data clustering solution with a number of clusters for unsupervised data. Silhouette can be used as valid index to decide the number of clusters and optimal solution by simultaneously considering intra- and inter-cluster distances. The performance of GSO using Silhouette is validated through several experiment and analysis of data sets.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• v.6 no.1
• /
• pp.10-14
• /
• 2006

Many clustering methods have been studied. For the most part of these methods may be needed to determine the number of clusters. But, there are few methods for determining the number of population clusters objectively. It is difficult to determine the cluster size. In general, the number of clusters is decided by subjectively prior knowledge. Because the results of clustering depend on the number of clusters, it must be determined seriously. In this paper, we propose an efficient method for determining the number of clusters using hybrid' self organizing map and new criterion for evaluating the clustering result. In the experiment, we verify our model to compare other clustering methods using the data sets from UCI machine learning repository.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.9 no.11
• /
• pp.1221-1226
• /
• 2014

In an effort to improve the efficiency of the speech recognition, we investigate the effect of the number of phoneme clusters. For this purpose, codebooks of varied number of phoneme clusters are prepared by modified k-means clustering algorithm. The subsequent processing is fuzzy vector quantization (FVQ) and hidden Markov model (HMM) for speech recognition test. The result shows that there are two distinct regimes. For large number of phoneme clusters, the recognition performance is roughly independent of it. For small number of phoneme clusters, however, the recognition error rate increases nonlinearly as it is decreased. From numerical calculation, it is found that this nonlinear regime might be modeled by a power law function. The result also shows that about 166 phoneme clusters would be the optimal number for recognition of 300 isolated words. This amounts to roughly 3 variations per phoneme.