• KIISE Transactions on Computing Practices
• /
• v.22 no.11
• /
• pp.614-618
• /
• 2016

As information on the internet and the data from smart devices are growing, the amount of stream data is also increasing in the real world. The stream data, which is a potentially large data, requires online learnable models and algorithms. In this paper, we propose a novel class of models: predictive convolutional neural networks to be able to perform online learning. These models are designed to deal with longer patterns as the layers become higher due to layering convolutional operations: detection and max-pooling on the time axis. As a preliminary check of the concept, we chose two-month gathered GPS data sequence as an observation sequence. On learning them with the proposed method, we compared the original sequence and the regenerated sequence from the abstract information of the models. The result shows that the models can encode long-range patterns, and can generate a raw observation sequence within a low error.

• Journal of radiological science and technology
• /
• v.39 no.2
• /
• pp.143-148
• /
• 2016

The purpose of this study was examined the measurement error factor on AMARES of jMRUI method for magnetic resonance spectroscopy (MRS) quantitative analysis by skilled and unskilled observer method and identified the reason of independent observers. The Point-resolved spectroscopy sequence was used to acquired magnetic resonance spectroscopy data of 10 weeks male Sprague-Dawley rat liver. The methylene protons ($(-CH_{2-})n$) of 1.3 ppm and water proton ($H_2O$) of 4.7 ppm ratio was calculated by LCModel software for using the reference data. The seven unskilled observers were calculated total lipid (methylene/water) using the jMRUI AMARES technique twice every 1 week, and we conducted interclass correlation coefficient (ICC) statistical analysis by SPSS software. The inter-observer reliability (ICC) of Cronbach's alpha value was less than 0.1. The average value of seven observer's total lipid ($0.096{\pm}0.038$) was 50% higher than LCModel reference value. The jMRUI AMARES analysis method is need to minimize the presence of the residual metabolite by identified metabolite MRS profile in order to obtain the same results as the LCModel.

• Journal of IKEEE
• /
• v.2 no.2 s.3
• /
• pp.212-224
• /
• 1998

There are two major factors of degrading the performance in the forward link of DS/CDMA systems. One is the multiple access interference (MAI) caused by using the same frequency bands simultaneously and the other is the multipath lading due to multipath propagation. PN codes which have minimum cross correlation properties among spread spectrum codes are necessary to reduce the MAI. In the conventional IS-95A system, the PN sequence has the period of $2^{15}$ and is of the length of 64 chips for spreading each data. In this case, since the length of PN code per bit is very short compared to the period of the PN code, the performance of the conventional system is not satisfied in view of suppressing the multipath interference. However, the correlation property of the PN codes at the demodulation can be improved by increasing the interval of Integration at the demodulation. This paper proposes a demodulation method to reduce the cross correlation among PN codes. The performance of the proposed demodulation method is investigated through computer simulations. We used multipath Ray lading channel and AWGN channel in the simulation. Our simulation results show the improved performance of $0.25{\sim}0.5dB$ SNR in a given BER compared to the conventional demodulation scheme.

• The KIPS Transactions:PartD
• /
• v.16D no.5
• /
• pp.661-672
• /
• 2009

Recently it was found that various genetic structural variations such as CNV(copy number variation) exist in the human genome, and these variations are closely related with disease susceptibility, reaction to treatment, and genetic characteristics. In this paper we propose a new CNV detection algorithm using millions of short DNA sequences generated by giga-sequencing technology. Our method maps the DNA sequences onto the reference sequence, and obtains the occurrence frequency of each read in the reference sequence. And then it detects the statistically significant regions which are longer than 1Kbp as the candidate CNV regions by analyzing the distribution of the occurrence frequency. To select a proper read alignment method, several methods are employed in our algorithm, and the performances are compared. To verify the superiority of our approach, we performed extensive experiments. The result of simulation experiments (using a reference sequence, build 35 of NCBI) revealed that our approach successfully finds all the CNV regions that have various shapes and arbitrary length (small, intermediate, or large size).

• Journal of the Korea Computer Graphics Society
• /
• v.15 no.4
• /
• pp.31-40
• /
• 2009

Captured motion data is widely used today in a variety of areas including film production, game development, sports analysis, and medical rehabilitation. The ability of analyzing and processing motion data has increased rapidly for last decades. However, it is still difficult for users to quickly understand the contents of motion data consisting of a series of time-varying poses. One typical approach is to visualize consecutive poses in sequence while adjusting three-dimensional view, which is often time-consuming and laborious especially when users need to repeatedly control time and view in order to search for desired motions. We present a method of visualizing motion data as a sequence of images in comic book layout so that users can rapidly understand the overall flows of motion data, and easily identify their desired motions. The usefulness of our approach is demonstrated by visualizing various kinds of motion data including locomotion, boxing, and interaction with environments.

• 한국HCI학회:학술대회논문집
• /
• 2008.02a
• /
• pp.744-748
• /
• 2008

In this paper, we propose a novel technique that generate a new song with given example songs. Our system use k-th order Markov chain of which each state represents notes in a measure. Because we have to consider very high-dimensional space if we use notes in a measure as a state of Markov chain directly, we exploit a hierarchical clustering technique for given example songs to use each cluster as a state. Each given examples can be represented as sequences of cluster ID, and we use them for training data of the Markov chain. The resulting Markov chain effectively gives new song similar to given examples.

• Journal of the Institute of Electronics Engineers of Korea TC
• /
• v.46 no.3
• /
• pp.26-31
• /
• 2009

In this paper, we propose the receiver algorithms suitable for the ECMA standard proposed for multi-gigabit packet transmission in 60 GHz band. In the ECMA standard, various modulation schemes are used for system flexibility. Hence, it is crucial to develop receiver algorithms supporting various modulation schemes with an uniform hardware structure. In this paper, we propose the receiver algorithms supporting DBPSK, DQPSK and OOK modulation schemes simultaneously. The proposed algorithms are not only hardware efficient but also support various modulation schemes with an uniform hardware structure.

• Proceedings of the Korean Society of Medical Physics Conference
• /
• 2005.04a
• /
• pp.93-95
• /
• 2005

MRS is to measure very small metabolite signals, whose resonant frequencies spread over the chemical shift range characteristic of the measured nucleus. The MR signal originates from the excited volume, which is a column of tissue divided into slices by gradient or rf encoding. The parameters that acquired data affected by TE, TR, and other variables. The higher spatial resolution of 3D CSI compared to SVS and its ability to examine regional metabolite variations for brain study.

• The Transactions of the Korea Information Processing Society
• /
• v.7 no.10
• /
• pp.3290-3299
• /
• 2000

Digital watermarking has similar concepts with channel coding thechnique for transferring data with minimizing error in noise environment, since it should be robust to various kinds of data manipulation for protecting copyrights of multimedia data. This paper proposes a digital watermarking technique which is robust to various kinds of data manipulation. Intellectual property rights information is encoded using a convolutional code, and block-interleaving technique is applied to prevent successive loss of encoded data. Encoded intelloctual property rithts informationis embedded using spread spectrum technique which is robust to cata manipulation. In order to reconstruct intellectual property rights information, watermark signalis detected by covariance between watermarked image and pseudo rando noise sequence which is used to einbed watermark. Embedded intellectual property rights information is obtaned by de-interleaving and cecoding previously detected wtermark signal. Experimental results show that block interleaving watermarking technique can detect embedded intellectial property right informationmore correctly against to attacks like Gaussian noise additon, filtering, and JPEG compression than general spread spectrum technique in the same PSNR.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.13 no.4
• /
• pp.805-812
• /
• 2009

In many intrusion detection applications, n-gram approach has been widely applied. However, n-gram approach has shown a few problems including unscalability and double counting of features. To address those problems, we applied n-gram augmented Naive Bayes with k-truncated suffix tree (k-TST) storage mechanism directly to classify intrusive sequences and compared performance with those of Naive Bayes and Support Vector Machines (SVM) with n-gram features by the experiments on host-based intrusion detection benchmark data sets. Experimental results on the University of New Mexico (UNM) benchmark data sets show that the n-gram augmented method, which solves the problem of independence violation that happens when n-gram features are directly applied to Naive Bayes (i.e. Naive Bayes with n-gram features), yields intrusion detectors with higher accuracy than those from Naive Bayes with n-gram features and shows comparable accuracy to those from SVM with n-gram features. For the scalable and efficient counting of n-gram features, we use k-truncated suffix tree mechanism for storing n-gram features. With the k-truncated suffix tree storage mechanism, we tested the performance of the classifiers up to 20-gram, which illustrates the scalability and accuracy of n-gram augmented Naive Bayes with k-truncated suffix tree storage mechanism.