• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.14 no.7
• /
• pp.2785-2799
• /
• 2020

We propose a novel speaker adaptation method using acoustic model clustering. The similarity of different speakers is defined by the cosine distance between their i-vectors (intermediate vectors), and various efficient clustering algorithms are applied to obtain a number of speaker subsets with different characteristics. The speaker-independent model is then retrained with the training data of the individual speaker subsets grouped by the clustering results, and an unknown speech is recognized by the retrained model of the closest cluster. The proposed method is applied to a large-scale speech recognition system implemented by a hybrid hidden Markov model and deep neural network framework. An experiment was conducted to evaluate the word error rates using Resource Management database. When the proposed speaker adaptation method using i-vector based clustering was applied, the performance, as compared to that of the conventional speaker-independent speech recognition model, was improved relatively by as much as 12.2% for the conventional fully neural network, and by as much as 10.5% for the bidirectional long short-term memory.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 1998.06b
• /
• pp.105-110
• /
• 1998

This paper proposes a method to manage individual information on large scale distributed geographic information systems. On such system, ordinary users usually cannot alter the contents of the server. The method proposed in this paper makes possible to alter the contents or add individual data onto such kinds of non-write-permitted data onto set. We call the method as GDSF, ‘geographic differential script file’. In this method, a client user makes a GDSF which contains the private information to be added onto the served data. Then, the client keeps the file on a local disk. After this, when the user uses the data, he applies the differential data sequence onto the down loaded data to restore the information. The GDSF is a collection of picture commands which tell pictures insertions, deletions, and modification operations. The GDSF also can contain the modification. The GDSF also can contain the modification of the attribute information of geographic entities. The method also applicable to modify data on a ROM device, for example CD-ROM or DVD-ROM. This paper describes the method and experimental results.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.13 no.10
• /
• pp.5244-5259
• /
• 2019

With the continuous development of LBS (Location Based Service) applications, privacy protection has become an urgent problem to be solved. Differential privacy technology is based on strict mathematical theory that provides strong privacy guarantees where it supposes that the attacker has the worst-case background knowledge and that knowledge has been applied to different research directions such as data query, release, and mining. The difficulty of this research is how to ensure data availability while protecting privacy. Spatial multidimensional data are usually released by partitioning the domain into disjointed subsets, then generating a hierarchical index. The traditional data-dependent partition methods need to allocate a part of the privacy budgets for the partitioning process and split the budget among all the steps, which is inefficient. To address such issues, a novel two-step partition algorithm is proposed. First, we partition the original dataset into fixed grids, inject noise and synthesize a dataset according to the noisy count. Second, we perform IH-Tree (Improved H-Tree) partition on the synthetic dataset and use the resulting partition keys to split the original dataset. The algorithm can save the privacy budget allocated to the partitioning process and obtain a more accurate release. The algorithm has been tested on three real-world datasets and compares the accuracy with the state-of-the-art algorithms. The experimental results show that the relative errors of the range query are considerably reduced, especially on the large scale dataset.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.8 no.11
• /
• pp.4170-4188
• /
• 2014

In this study, we present a denoising algorithm for high-frame-rate videos in an ultra-low illumination environment on the basis of Kalman filtering model and a new motion segmentation scheme. The Kalman filter removes temporal noise from signals by propagating error covariance statistics. Regarded as the process noise for imaging, motion is important in Kalman filtering. We propose a new motion estimation scheme that is suitable for serious noise. This scheme employs the small motion vector characteristic of high-frame-rate videos. Small changing patches are intentionally neglected because distinguishing details from large-scale noise is difficult and unimportant. Finally, a spatial bilateral filter is used to improve denoising capability in the motion area. Experiments are performed on videos with both synthetic and real noises. Results show that the proposed algorithm outperforms other state-of-the-art methods in both peak signal-to-noise ratio objective evaluation and visual quality.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.12 no.9
• /
• pp.4429-4447
• /
• 2018

Local binary descriptors are well-suited for many real-time and/or large-scale computer vision applications, while their low computational complexity is usually accompanied by the limitation of performance. In this paper, we propose a new optimization framework, RLDB (Robust-LDB), to improve a typical region-based binary descriptor LDB (local difference binary) and maintain its computational simplicity. RLDB extends the multi-feature strategy of LDB and applies a more complete region-comparing configuration. A cascade bit selection method is utilized to select the more representative patterns from massive comparison pairs and an online learning strategy further optimizes descriptor for each specific patch separately. They both incorporate LDP (linear discriminant projections) principle to jointly guarantee the robustness and distinctiveness of the features from various scales. Experimental results demonstrate that this integrated learning framework significantly enhances LDB. The improved descriptor achieves a performance comparable to floating-point descriptors on many benchmarks and retains a high computing speed similar to most binary descriptors, which better satisfies the demands of applications.

• Nuclear Engineering and Technology
• /
• v.27 no.5
• /
• pp.670-686
• /
• 1995

This paper provides an improvement on our previous study ［1］ for multi-fault diagnosis in real time in large-scale systems. In the method, fault propagation probability(FPP) and fault propagation time(FPT) in a fuzzy sense are additively used to describe the fault propagation model(FPM) in more practical manner. A modified fault diagnosis procedure is also given. This method is applied for diagnosis of the primary system in the Kori nuclear power plant unit 2 under a transient condition in case of unit value of FPP on each branch of the FPM.

• Proceedings of the KSR Conference
• /
• 2008.06a
• /
• pp.188-193
• /
• 2008

According to introduce KTX in Korea, rail-road bridge section of KTX was increased approximately 50% of the total length. Bridge is required periodic inspection and check to prevent accident and hazard because various damage which have effects on traffic and replacement of damaged parts is difficult. Specifically, the train as large-scale transportation because accidents led to great damage, preventing these accidents are critical. Well-organized management and maintenance systems are required to prevent the accidents. In the case of roadway bridge, bridge inspection vehicle is used to deploy inspectors in roadway bridge. However, this method requires a lot of time and efforts, and inspectors are exposed to potential hazard. Also, surrounding environment like poor lighting system or electric wire could harm the inspector while repairing. Due to this reason, automatic repairing and inspecting system have been introduced to replace the old methods. Management system of the railroad bridge track for trains uses various advanced equipments, but whereas roadway bridge management system is lacking these efforts. As a result of that, this study looks over the existing management method. and review the method to apply the Bridge Inspection Robot in railroad bridge. Moreover, this study suggests future management technology using inspection robot.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2015.10a
• /
• pp.651-653
• /
• 2015

Common Embedded boards like the Arduino, Raspberry Pi, BeagleBone Black, leverages smart home systems, machine tools and various products in our day to day life. Managing and dealing frequent large scale incidents involving recent reagents and hazardous materials can be dangerous and difficult to detect in advance like in an event of an accidents or fires. In this paper I have done research by utilizing an Embedded (BeagleBone Black) boards sensors and RFID management system to detect a hazardous situation like fire in real time and avoiding it by sending out an alert message to the admin user to minimizing the risk. This system provides immediate information to the administrator of any hazardous situation and prevents any accidents from occurring.

• Corrosion Science and Technology
• /
• v.15 no.5
• /
• pp.245-252
• /
• 2016

The pipelines and equipments are degraded by flow-accelerated corrosion (FAC), and a large-scale test facility was constructed for simulate the FAC phenomena in secondary coolant environment of PWR type nuclear power plants. Using this facility, FAC test was performed on weld pipe (carbon steel and low alloy steel) at the conditions of high velocity flow (> 10 m/s). Wall thickness was measured by high temperature ultrasonic monitoring systems (four-channel buffer rod type and waveguide type) during test period and room temperature manual ultrasonic method before and after test period. This work deals with the complex effects of flow velocity on the wall thinning in weld pipe and the test results showed that the higher flow velocity induced different increasement of wall thinning rate for the carbon steel and low alloy steel pipe.

• Journal of Communications and Networks
• /
• v.15 no.3
• /
• pp.283-291
• /
• 2013

In this paper, we propose new multiuser detectors (MUDs) based on compressed sensing approaches for the large-scale multiple antenna systems equipped with dozens of low-power antennas. We consider the scenarios where the number of receiver antennas is smaller than the total number of users, but the number of active users is relatively small. This prior information motivates sparsity-embracing MUDs such as sparsity-embracing linear/nonlinear MUDs where the detection of active users and their symbol detection are employed. In addition, sparsity-embracing MUDs with maximum a posteriori probability criterion (MAP-MUDs) are presented. They jointly detect active users and their symbols by exploiting the probability of user activity, and it can be solved efficiently by introducing convex relaxing senses. Furthermore, it is shown that sparsity-embracing MUDs exploiting common users' activity across multiple symbols, i.e., frame-by-frame, can be considered to improve performance. Also, in multiple multiple-input and multiple-output networks with aggressive frequency reuse, we propose the interference cancellation strategy for the proposed sparsity-embracing MUDs. That first cancels out the interference induced by adjacent networks and then recovers the desired users' information by exploiting the low user activity. In simulation studies for binary phase shift keying modulation, numerical evidences establish the effectiveness of our proposed MUDs exploiting low user activity, as compared with the conventional MUD.