• The Journal of the Acoustical Society of Korea
• /
• v.13 no.3
• /
• pp.63-70
• /
• 1994

In this Paper, we propose a new codeword tying algorithm based on a tree structured classfier. The proposed algorithm which can be viewed as a kind of soft decision using statistical properties between codewords and states has an advantage of fast construction, and guarantees a unique optimal solution. Also, it can easily be applied to any speech recognition system based on discrete hidden Markov model (HMM). Experimental results on speaker-independent isolated word recognition show error reduction of $6\%$ for the codebook of size 256 and $9\%$ for 512 size and also HMM parameter reduction of about $20\%$.

• The KIPS Transactions:PartB
• /
• v.18B no.2
• /
• pp.87-92
• /
• 2011

In Korean documents, there are diverse spellings of transliterated foreign loanwords. This fact diminishes the performance of information retrieval systems in that a foreign word can be recognized differently, which is to say, as two or several different words. Thus, information retrieval systems require preprocessing to correct nonstandard loanword spellings prior to searching and recognizing corresponding equivalent words. This paper proposes a method that improves precision and processing efficiency using the Korean alphabet's viable prefix, which prunes a virtual tree from which candidate loanwords are created.

• The Journal of the Acoustical Society of Korea
• /
• v.22 no.7
• /
• pp.539-544
• /
• 2003

This paper describes how to reduce the effect of an occupation threshold by that the transform of mixture components of HMM parameters is controlled in hierarchical tree structure to prevent from over-adaptation. To reduce correlations between data elements and to remove elements with less variance, we employ PCA (Principal component analysis) and ICA (independent component analysis) that would give as good a representation as possible, and decline the effect of over-adaptation. When we set lower occupation threshold and increase the number of transformation function, ordinary MLLR adaptation algorithm represents lower recognition rate than SI models, whereas the proposed MLLR adaptation algorithm represents the improvement of over 2% for the word recognition rate as compared to performance of SI models.

• Proceedings of the IEEK Conference
• /
• 2003.07a
• /
• pp.214-217
• /
• 2003

In this paper, we deal with implementing design for a correlator access code generator module which they are used for setting up a connection between units, a packet decision, a clock syncronization, by FPGA. The orrelator module which is composed of the Wallace Tree's CSA and threshold value decision device decides useful a packet and syncronizes a clock, after it correlates an input signal of 1 Mbps transmission rate by a sliding window. An access code generator module which is composed of a BCH (Bose-Chadhuri-Hocquenghem) cyclic encoder and control device was designed according as a four steps' generation process proposed in the bluetooth standard. The pseudo random sequence which solves syncronization problem saved a voluntary device Proposed the module was designed by VHDL. An simulation and test are inspected by Xilinx FPGA.

• Journal of the Institute of Electronics Engineers of Korea SD
• /
• v.41 no.4
• /
• pp.57-66
• /
• 2004

This paper proposes the design of a 2,048-bit RSA based on RNS(residue number systems) Montgomery modular multiplier As the systems that RNS processes a fast parallel modular multiplication for a large word partitioned into small words, we introduce Montgomery reduction method(MRM)［1］based on Wallace tree modular multiplier and 33 RNS bases with 64-bit size for RNS Montgomery modular multiplication in this paper. Also, for fast RNS modular multiplication, a modified method based on Chinese remainder theorem(CRT)［2］ is presented. We have verified 2,048-bit RSA based on RNS using Samsung 0.35${\mu}{\textrm}{m}$ technology and the 2,048-bit RSA is performed in 2.54㎳ at 100MHz.

• Journal of Digital Contents Society
• /
• v.6 no.1
• /
• pp.89-94
• /
• 2005

In machine translation the goal of natural language generation is to produce an target sentence transmitting the meaning of source sentence by using an parsing tree of source sentence and target expressions. It provides generator with linguistic structures, word mapping, part-of-speech, lexical information. The purpose of this study is to research the Korean Characteristics which could be used for the establishment of an algorism in speech recognition and composite sound. This is a part of realization for the plan of automatic machine translation. The stage of MT is divided into the level of morphemic, semantic analysis and syntactic construction.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.9 no.1
• /
• pp.391-406
• /
• 2015

Social media have emerged as new communication channels between consumers and companies that generate a large volume of unstructured text data. This social media content, which contains consumers' opinions and interests, is recognized as valuable material from which businesses can mine useful information; consequently, many researchers have reported on opinion-mining frameworks, methods, techniques, and tools for business intelligence over various industries. These studies sometimes focused on how to use opinion mining in business fields or emphasized methods of analyzing content to achieve results that are more accurate. They also considered how to visualize the results to ensure easier understanding. However, we found that such approaches are often technically complex and insufficiently user-friendly to help with business decisions and planning. Therefore, in this study we attempt to formulate a more comprehensive and practical methodology to conduct social media opinion mining and apply our methodology to a case study of the oldest instant noodle product in Korea. We also present graphical tools and visualized outputs that include volume and sentiment graphs, time-series graphs, a topic word cloud, a heat map, and a valence tree map with a classification. Our resources are from public-domain social media content such as blogs, forum messages, and news articles that we analyze with natural language processing, statistics, and graphics packages in the freeware R project environment. We believe our methodology and visualization outputs can provide a practical and reliable guide for immediate use, not just in the food industry but other industries as well.

• Speech Sciences
• /
• v.8 no.4
• /
• pp.17-33
• /
• 2001

In this paper, we implement the large variable vocabulary speech recognition system, which is characterized by no additional pre-training process and no limitation of recognized word list. We have designed the system in order to achieve the high recognition rate using the decision tree based state tying algorithm and in order to reduce the processing time using the gaussian selection based variable flooring algorithm, the limitation algorithm of the number of nodes and ENNS algorithm. The gaussian selection based variable flooring algorithm shows that it can reduce the total processing time by more than half of the recognition time, but it brings about the reduction of recognition rate. In other words, there is a trade off between the recognition rate and the processing time. The limitation algorithm of the number of nodes shows the best performance when the number of gaussian mixtures is a three. Both of the off-line and on-line experiments show the same performance. In our experiments, there are some differences of the recognition rate and the average recognition time according to the distinction of genders, speakers, and the number of vocabulary.

• The KIPS Transactions:PartD
• /
• v.8D no.5
• /
• pp.587-592
• /
• 2001

Many applications are being developed to automate office works, and the electronic dictionary(e-Dictionary) is one of the main components of the office suites. Several requirements are proposed for the efficient e-dictionaries :1) Fast searching time, 2) Data compatibility with other e-dictionaries to deal with words and obsolete word, and 3) Reusable components to develop new customized e-dictionaries with minimized development time and cost. We propose a data format with which any e-dictionary can change data with others. We also develop System Dictionary component and Customer Dictionary component to enable-and-play component reuse. Our e-dictionary achieves fast searching time by efficiently managing Trie and B-tree index structure for the dictionary components.

• The Transactions of the Korean Institute of Electrical Engineers D
• /
• v.53 no.3
• /
• pp.197-205
• /
• 2004

In this paper, we studied on the hardware implementation of a 3${\times}$3 window weighted median filter using bit-level sorting algorithm. The weighted median filter is a generalization of the median filter that is able to preserve :,harp changes in signal and is very effective in removing impulse noise. It has been successfully applied in various areas such as digital signal and video/image processing. The weighted median filters are, for the most part, based on word-level sorting methods, which have more hardware and time complexity, However, the proposed bit-serial sorting algorithm uses weighted adder tree to overcome those disadvantages. It also offers a simple pipelined filter architecture that is highly regular with repeated modules and is very suitable for weighted median filtering. The algorithm was implemented by VHDL and graphical environment in MAX＋PlusII of ALTERA. The simulation results indicate that the proposed design method is more efficient than the traditional ones.