• MALSORI
• /
• no.59
• /
• pp.53-68
• /
• 2006

We report the evaluation results of the Korean speech recognition platform called ECHOS. The platform has an object-oriented and reusable architecture so that researchers can easily evaluate their own algorithms. The platform has all intrinsic modules to build a large vocabulary speech recognizer: Noise reduction, end-point detection, feature extraction, hidden Markov model (HMM)-based acoustic modeling, cross-word modeling, n-gram language modeling, n-best search, word graph generation, and Korean-specific language processing. The platform supports both lexical search trees and finite-state networks. It performs word-dependent n-best search with bigram in the forward search stage, and rescores the lattice with trigram in the backward stage. In an 8000-word continuous speech recognition task, the platform with a lexical tree increases 40% of word errors but decreases 50% of recognition time compared to the HTK platform with flat lexicon. ECHOS reduces 40% of recognition errors through incorporation of cross-word modeling. With the number of Gaussian mixtures increasing to 16, it yields word accuracy comparable to the previous lexical tree-based platform, Julius.

• Journal of Information Processing Systems
• /
• v.19 no.2
• /
• pp.139-148
• /
• 2023

Current food-traceability platforms suffer from problems such as inconsistent traceability standards, a lack of public credibility, and slow access to data. In this work, a combined code and identification method was designed that can achieve more secure product traceability using the dual anti-counterfeiting technology of a QR code and a hidden code. When the QR code is blurry, the hidden code can still be used to effectively identify food information. Based on this combined code, a food-safety traceability platform was developed. The platform follows unified encoding standards and provides standardized interfaces. Based on this innovation, the platform not only can serve individual food-traceability systems development, but also connect existing traceability systems. These will help to solve the problems such as non-standard traceability content, inconsistent processes, and incompatible system software. The experimental results show that the combined code has higher accuracy. The food-safety traceability platform based on the combined code improves the safety of the traceability process and the integrity of the traceability information. The innovation of this paper is invoking the combined code united the QR code's rapidity and the hidden code's reliability, developing a platform that uses a unified coding standard and provides a standardized interface to resolve the differences between multi-food-traceability systems. Among similar systems, it is the only one that has been connected to the national QR code identification platform. The project has made profits and has significant economic and social benefits.

• MALSORI
• /
• no.51
• /
• pp.151-165
• /
• 2004

For educational and research purposes, a Korean speech recognition platform is designed. It is based on an object-oriented architecture and can be easily modified so that researchers can readily evaluate the performance of a recognition algorithm of interest. This platform will save development time for many who are interested in speech recognition. The platform includes the following modules: Noise reduction, end-point detection, met-frequency cepstral coefficient (MFCC) and perceptually linear prediction (PLP)-based feature extraction, hidden Markov model (HMM)-based acoustic modeling, n-gram language modeling, n-best search, and Korean language processing. The decoder of the platform can handle both lexical search trees for large vocabulary speech recognition and finite-state networks for small-to-medium vocabulary speech recognition. It performs word-dependent n-best search algorithm with a bigram language model in the first forward search stage and then extracts a word lattice and restores each lattice path with a trigram language model in the second stage.

• Preventive Nutrition and Food Science
• /
• v.19 no.4
• /
• pp.343-347
• /
• 2014

The present study investigated the effects of Vaccinium uliginosum L. (bilberry) on the learning and memory impairments induced by amyloid-${\beta}$ protein ($A{\beta}P$) 1-42. ICR Swiss mice were divided into 4 groups: the control ($A{\beta}40$-1A), control with 5% bilberry group ($A{\beta}40$-1B), amyloid ${\beta}$ protein 1-42 treated group ($A{\beta}1$-42A), and $A{\beta}1$-42 with 5% bilberry group ($A{\beta}1$-42B). The control was treated with amyloid ${\beta}$-protein 40-1 for placebo effect, and Alzheimer's disease (AD) group was treated with amyloid ${\beta}$-protein 1-42. Amyloid ${\beta}$-protein 1-42 was intracerebroventricular (ICV) micro injected into the hippocampus in 35% acetonitrile and 0.1% trifluoroacetic acid. Although bilberry added groups tended to decrease the finding time of hidden platform, no statistical significance was found. On the other hand, escape latencies of $A{\beta}P$ injected mice were extended compared to that of $A{\beta}40$-1. In the Probe test, bilberry added $A{\beta}1$-42B group showed a significant (P<0.05) increase of probe crossing frequency compared to $A{\beta}1$-42A. Administration of amyloid protein ($A{\beta}1$-42) decreased working memory compared to $A{\beta}40$-1 control group. In passive avoidance test, bilberry significantly (P<0.05) increased the time of staying in the lighted area compared to AD control. The results suggest that bilberry may help to improve memory and learning capability in chemically induced Alzheimer's disease in experimental animal models.

• Journal of Internet Computing and Services
• /
• v.20 no.6
• /
• pp.73-84
• /
• 2019

Machine learning as a service, the so-called MLaaS, has recently attracted much attention in almost all industries and research groups. The main reason for this is that you do not need network servers, storage, or even data scientists, except for the data itself, to build a productive service model. However, machine learning is often very difficult for most developers, especially in traditional science due to the lack of well-structured big data for scientific data. For experiment or application researchers, the results of an experiment are rarely shared with other researchers, so creating big data in specific research areas is also a big challenge. In this paper, we introduce the KISTI-ML platform, a community-based rapid AI model development for scientific data. It is a place where machine learning beginners use their own data to automatically generate code by providing a user-friendly online development environment. Users can share datasets and their Jupyter interactive notebooks among authorized community members, including know-how such as data preprocessing to extract features, hidden network design, and other engineering techniques.

• Smart Media Journal
• /
• v.1 no.4
• /
• pp.18-26
• /
• 2012

The increase of the number of smartphone applications has enforced the importance of new user interface emergence and has raised the interest of research in the convergence of multiple sensors. In this paper, we propose a method for the convergence of acceleration, magnetic and gyro sensors to recognize the gesture from motion of user smartphone. The proposed method first obtain the 3D orientation of smartphone and recognize the gesture of hand motion by using HMM(Hidden Markov Model). The proposed method for the representation for 3D orientation of smartphone in spherical coordinate was used for quantization of smartphone orientation to be more sensitive in rotation axis. The experimental result shows that the success rate of our method is 93%.

• Journal of the Korea Society of Computer and Information
• /
• v.25 no.5
• /
• pp.27-36
• /
• 2020

In this paper, we propose a method to derive valuable but hidden infromation from the data which is the core foundation in the 4th Industrial Revolution to pursue knowledge-based service fusion. The hyper-connected societies characterized by IoT inevitably produce big data, and with the data in order to derive optimal services for trouble situations it is first processed by discovering valuable information. A data-centric IoT platform is a platform to collect, store, manage, and integrate the data from variable devices, which is actually a type of middleware platforms. Its purpose is to provide suitable solutions for challenged problems after processing and analyzing the data, that depends on efficient and accurate algorithms performing the work of data analysis. To this end, we propose specially designed structures to store IoT data without losing the semantics and provide algorithms to discover the useful information with several definitions and proofs to show the soundness.

• Journal of Nutrition and Health
• /
• v.39 no.3
• /
• pp.219-224
• /
• 2006

This study was performed to determine the effect of soy isoflavones on brain development and function in rats. Forty Sprague-Dawley male rats were provided diets containing different levels of soy isoflavones for 6 weeks; 0 ppm (control), 50 ppm (low isoflavone intake; LI), 250 ppm (medium isoflavone intake; MI) and 500 ppm (high isoflavone intake; HI). Learning ability was evaluated by a Y-shaped water maze and the activity of acetylcholinesterase in brain was assayed after decapitation. Food intake and body weights as well as weights of brain, liver, spleen, heart and kidney showed no significant difference among the four groups, which means 500 ppm of isoflavones is safe. In the water maze test, the frequency of error counted when rats entered one end of the alley without platform was significantly lower in the HI group than in the control group, and the escape latency as swim time taken to escape on the hidden platform was significantly shorter in the HI group than in the LI and control groups. The activity of acetylcholinesterase of the brain was significantly higher in the HI and MI groups than in the control group. Therefore, the results indicate that isoflavones may improve the cognitive function without adverse effects.

• Journal of Ginseng Research
• /
• v.20 no.2
• /
• pp.139-148
• /
• 1996

This study was designed to examine the effects of red ginseng total saponin and extract on spatial working memory in mice using Morris water maze. Two kinds of red ginseng saponin (No. 1 and No. 2) and three kinds of red ginseng extract (No. 1, No. 2 and No. 3) to have different PD/ PT ratio (No. 1=1.24, No.2=1.47 No.3=2.41) were prepared by mixing the different parts of red ginseng In different ratio. In acute administration of total saponin No. 1 or No. 2, escape time to reach to a hidden platform In a fixed location for training trials was significantly decreased as compared with control group and swimming time in the quadrant that had contained the platform was also significantly increased as compared with control group. In acute treatment of extract No. 1 or 1 No. 2, swimming time in the platformless quadrant was increased dose dependently as compared with control group, especially at dose of 200 mg/kg,bw swimming time was significantly Increased. Oral treatment of extract No. 1 (100 mg/kg, bw) for 7 days produced an increase of swimming time In the platformless quadrant but a decrease of swimming time in No.3-treated group (100 mg/kg, bw). These results show that red ginseng may improve spatial discrimination learning and spatial working memory of mice

• Journal of Biomedical Engineering Research
• /
• v.42 no.5
• /
• pp.232-240
• /
• 2021

Heterogeneity in cancer is the major obstacle for precision medicine and has become a critical issue in the field of a cancer diagnosis. Many attempts were made to disentangle the complexity by molecular classification. However, multi-dimensional information from dynamic responses of cancer poses fundamental limitations on biomolecular marker-based conventional approaches. Cell morphology, which reflects the physiological state of the cell, can be used to track the temporal behavior of cancer cells conveniently. Here, we first present a hybrid learning-based platform that extracts cell morphology in a time-dependent manner using a deep convolutional neural network to incorporate multivariate data. Feature selection from more than 200 morphological features is conducted, which filters out less significant variables to enhance interpretation. Our platform then performs unsupervised clustering to unveil dynamic behavior patterns hidden from a high-dimensional dataset. As a result, we visualize morphology state-space by two-dimensional embedding as well as representative morphology clusters and trajectories. This cell morphology profiling strategy by hybrid learning enables simplification of the heterogeneous population of cancer.