• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.19 no.1
• /
• pp.178-184
• /
• 2015

This paper proposes a hardware architecture for high performance Deblocking filter(DBF) in High Efficiency Video Coding for UHD(Ultra High Definition) videos. This proposed hardware architecture which has less processing time has a 4-stage pipelined architecture with two filters and parallel boundary strength module. Also, the proposed filter can be used in low-voltage design by using clock gating architecture in 4-stage pipeline. The segmented memory architecture solves the hazard issue that arises when single port SRAM is accessed. The proposed order of filtering shortens the delay time that arises when storing data into the single port SRAM at the pre-processing stage. The DBF hardware proposed in this paper was designed with Verilog HDL, and was implemented with 22k logic gates as a result of synthesis using TSMC 0.18um CMOS standard cell library. Furthermore, the dynamic frequency can process UHD 8k($7680{\times}4320$) samples@60fps using a frequency of 150MHz with an 8K resolution and maximum dynamic frequency is 285MHz. Result from analysis shows that the proposed DBF hardware architecture operation cycle for one process coding unit has improved by 32% over the previous one.

• Journal of the Korea Society of Computer and Information
• /
• v.18 no.11
• /
• pp.1-12
• /
• 2013

In multi-core processors, Last Level Cache(LLC) can reduce the speed gap between the memory and the core. For this reason, LLC has big impact on the performance of processors. LLC is composed of shared cache and private cache. In computer architecture community, most researchers have mainly focused on the management techniques for shared cache, while management techniques for private cache have not been widely researched. In conventional private LLC, memory is statically assigned to each core, resulting in serious performance degradation when the workloads are not fairly distributed. To overcome this problem, this paper proposes the replacement policy for managing private cache of LLC efficiently. As proposed core-aware cache replacement policy can reconfigure LLC dynamically, hit rate of LLC is increases drastically. Moreover, proposed policy uses 2-bit saturating counters to improve the performance. According to our simulation results, the proposed method can improve hit rates by 9.23% and reduce the access time by 12.85% compared to the conventional method.

• 재활복지
• /
• v.21 no.3
• /
• pp.129-146
• /
• 2017

In an event of an emergency, such as fire in a building, visually impaired and blind people are prone to exposed to a level of danger that is greater than that of normal people, for they cannot be aware of it quickly. Current fire detection methods such as smoke detector is very slow and unreliable because it usually uses chemical sensor based technology to detect fire particles. But by using vision sensor instead, fire can be proven to be detected much faster as we show in our experiments. Previous studies have applied various image processing and machine learning techniques to detect fire, but they usually don't work very well because these techniques require hand-crafted features that do not generalize well to various scenarios. But with the help of recent advancement in the field of deep learning, this research can be conducted to help solve this problem by using deep learning-based object detector that can detect fire using images from security camera. Deep learning based approach can learn features automatically so they can usually generalize well to various scenes. In order to ensure maximum capacity, we applied the latest technologies in the field of computer vision such as YOLO detector in order to solve this task. Considering the trade-off between recall vs. complexity, we introduced two convolutional neural networks with slightly different model's complexity to detect fire at different recall rate. Both models can detect fire at 99% average precision, but one model has 76% recall at 30 FPS while another has 61% recall at 50 FPS. We also compare our model memory consumption with each other and show our models robustness by testing on various real-world scenarios.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2014.10a
• /
• pp.271-274
• /
• 2014

This paper proposes high-performance SAO(Sample Adaptive Offset) in HEVC(High Efficiency Video Coding) encoder for Ultra HD video processing in real time. SAO is a newly adopted technique belonging to the in-loop filter in HEVC. The proposed SAO encoder hardware architecture uses three-layered buffers to minimize memory access time and to simplify pixel processing and also uses only adder, subtractor, shift register and feed-back comparator to reduce area. Furthermore, the proposed architecture consists of pipelined pixel classification and applying SAO parameters, and also classifies four consecutive pixels into EO and BO concurrently. These result in the reduction of processing time and computation. The proposed SAO encoder architecture is designed by Verilog HDL, and implemented by 180k logic gates in TSMC $0.18{\mu}m$ process. At 110MHz, the proposed SAO encoder can support 4K Ultra HD video encoding at 30fps in real time.

• Journal of Korean Society of Disaster and Security
• /
• v.14 no.4
• /
• pp.17-27
• /
• 2021

Streamflow prediction is a very vital disaster mitigation approach for effective flood management and water resources planning. Lately, torrential rainfall caused by climate change has been reported to have increased globally, thereby causing enormous infrastructural loss, properties and lives. This study evaluates the contribution of rainfall to streamflow prediction in normal and peak rainfall scenarios, typical of the recent flood at Piney Resort in Vernon, Hickman County, Tennessee, United States. Daily streamflow, water level, and rainfall data for 20 years (2000-2019) from two USGS gage stations (03602500 upstream and 03599500 downstream) of the Piney River watershed were obtained, preprocesssed and fitted with Long short term memory (LSTM) model. Tensorflow and Keras machine learning frameworks were used with Python to predict streamflow values with a sequence size of 14 days, to determine whether the model could have predicted the flooding event in August 21, 2021. Model skill analysis showed that LSTM model with full data (water level, streamflow and rainfall) performed better than the Naive Model except some rainfall models, indicating that only rainfall is insufficient for streamflow prediction. The final LSTM model recorded optimal NSE and RMSE values of 0.68 and 13.84 m³/s and predicted peak flow with the lowest prediction error of 11.6%, indicating that the final model could have predicted the flood on August 24, 2021 given a peak rainfall scenario. Adequate knowledge of rainfall patterns will guide hydrologists and disaster prevention managers in designing efficient early warning systems and policies aimed at mitigating flood risks.

• KIPS Transactions on Software and Data Engineering
• /
• v.12 no.4
• /
• pp.179-188
• /
• 2023

Recently, fake news disguises the form of news content and appears whenever important events occur, causing social confusion. Accordingly, artificial intelligence technology is used as a research to detect fake news. Fake news detection approaches such as automatically recognizing and blocking fake news through natural language processing or detecting social media influencer accounts that spread false information by combining with network causal inference could be implemented through deep learning. However, fake news detection is classified as a difficult problem to solve among many natural language processing fields. Due to the variety of forms and expressions of fake news, the difficulty of feature extraction is high, and there are various limitations, such as that one feature may have different meanings depending on the category to which the news belongs. In this paper, emotional change patterns are presented as an additional identification criterion for detecting fake news. We propose a model with improved performance by applying a convolutional neural network to a fake news data set to perform analysis based on content characteristics and additionally analyze emotional change patterns. Sentimental polarity is calculated for the sentences constituting the news and the result value dependent on the sentence order can be obtained by applying long-term and short-term memory. This is defined as a pattern of emotional change and combined with the content characteristics of news to be used as an independent variable in the proposed model for fake news detection. We train the proposed model and comparison model by deep learning and conduct an experiment using a fake news data set to confirm that emotion change patterns can improve fake news detection performance.

• Journal of Internet Computing and Services
• /
• v.16 no.2
• /
• pp.77-83
• /
• 2015

Frequent pattern mining, which is one of the major areas actively studied in data mining, is a method for extracting useful pattern information hidden from large data sets or databases. Moreover, frequent pattern mining approaches have been actively employed in a variety of application fields because the results obtained from them can allow us to analyze various, important characteristics within databases more easily and automatically. However, traditional frequent pattern mining methods, which simply extract all of the possible frequent patterns such that each of their support values is not smaller than a user-given minimum support threshold, have the following problems. First, traditional approaches have to generate a numerous number of patterns according to the features of a given database and the degree of threshold settings, and the number can also increase in geometrical progression. In addition, such works also cause waste of runtime and memory resources. Furthermore, the pattern results excessively generated from the methods also lead to troubles of pattern analysis for the mining results. In order to solve such issues of previous traditional frequent pattern mining approaches, the concept of representative pattern mining and its various related works have been proposed. In contrast to the traditional ones that find all the possible frequent patterns from databases, representative pattern mining approaches selectively extract a smaller number of patterns that represent general frequent patterns. In this paper, we describe details and characteristics of pattern condensing techniques that consider the maximality or closure property of generated frequent patterns, and conduct comparison and analysis for the techniques. Given a frequent pattern, satisfying the maximality for the pattern signifies that all of the possible super sets of the pattern must have smaller support values than a user-specific minimum support threshold; meanwhile, satisfying the closure property for the pattern means that there is no superset of which the support is equal to that of the pattern with respect to all the possible super sets. By mining maximal frequent patterns or closed frequent ones, we can achieve effective pattern compression and also perform mining operations with much smaller time and space resources. In addition, compressed patterns can be converted into the original frequent pattern forms again if necessary; especially, the closed frequent pattern notation has the ability to convert representative patterns into the original ones again without any information loss. That is, we can obtain a complete set of original frequent patterns from closed frequent ones. Although the maximal frequent pattern notation does not guarantee a complete recovery rate in the process of pattern conversion, it has an advantage that can extract a smaller number of representative patterns more quickly compared to the closed frequent pattern notation. In this paper, we show the performance results and characteristics of the aforementioned techniques in terms of pattern generation, runtime, and memory usage by conducting performance evaluation with respect to various real data sets collected from the real world. For more exact comparison, we also employ the algorithms implementing these techniques on the same platform and Implementation level.

• Journal of Internet Computing and Services
• /
• v.14 no.6
• /
• pp.1-8
• /
• 2013

Data mining techniques are used to find important and meaningful information from huge databases, and pattern mining is one of the significant data mining techniques. Pattern mining is a method of discovering useful patterns from the huge databases. Frequent pattern mining which is one of the pattern mining extracts patterns having higher frequencies than a minimum support threshold from databases, and the patterns are called frequent patterns. Traditional frequent pattern mining is based on a single minimum support threshold for the whole database to perform mining frequent patterns. This single support model implicitly supposes that all of the items in the database have the same nature. In real world applications, however, each item in databases can have relative characteristics, and thus an appropriate pattern mining technique which reflects the characteristics is required. In the framework of frequent pattern mining, where the natures of items are not considered, it needs to set the single minimum support threshold to a too low value for mining patterns containing rare items. It leads to too many patterns including meaningless items though. In contrast, we cannot mine any pattern if a too high threshold is used. This dilemma is called the rare item problem. To solve this problem, the initial researches proposed approximate approaches which split data into several groups according to item frequencies or group related rare items. However, these methods cannot find all of the frequent patterns including rare frequent patterns due to being based on approximate techniques. Hence, pattern mining model with multiple minimum supports is proposed in order to solve the rare item problem. In the model, each item has a corresponding minimum support threshold, called MIS (Minimum Item Support), and it is calculated based on item frequencies in databases. The multiple minimum supports model finds all of the rare frequent patterns without generating meaningless patterns and losing significant patterns by applying the MIS. Meanwhile, candidate patterns are extracted during a process of mining frequent patterns, and the only single minimum support is compared with frequencies of the candidate patterns in the single minimum support model. Therefore, the characteristics of items consist of the candidate patterns are not reflected. In addition, the rare item problem occurs in the model. In order to address this issue in the multiple minimum supports model, the minimum MIS value among all of the values of items in a candidate pattern is used as a minimum support threshold with respect to the candidate pattern for considering its characteristics. For efficiently mining frequent patterns including rare frequent patterns by adopting the above concept, tree based algorithms of the multiple minimum supports model sort items in a tree according to MIS descending order in contrast to those of the single minimum support model, where the items are ordered in frequency descending order. In this paper, we study the characteristics of the frequent pattern mining based on multiple minimum supports and conduct performance evaluation with a general frequent pattern mining algorithm in terms of runtime, memory usage, and scalability. Experimental results show that the multiple minimum supports based algorithm outperforms the single minimum support based one and demands more memory usage for MIS information. Moreover, the compared algorithms have a good scalability in the results.

• Journal of KIISE:Computer Systems and Theory
• /
• v.34 no.5_6
• /
• pp.257-266
• /
• 2007

In On-Chip bus, the arbitration scheme is one of the critical factors that decide the overall system performance. The arbitration scheme used in traditional shared bus is the master-side arbitration based on the request and grant signals between multiple masters and single arbiter. In the case of the master-side arbitration, only one master and one slave can transfer the data at a time. Therefore the throughput of total bus system and the utilization of resources are decreased in the master-side arbitration. However in the slave-side arbitration, there is an arbiter at each slave port and the master just starts a transaction and waits for the slave response to proceed to the next transfer. Thus, the unit of arbitration can be a transaction or a transfer. Besides the throughput of total bus system and the utilization of resources are increased since the multiple masters can simultaneously perform transfers with independent slaves. In this paper, we implement and analyze the arbitration schemes for the Multi-Layer AHB BusMatrix based on the slave-side arbitration. We implement the slave-side arbitration schemes based on fixed priority, round robin and dynamic priority and accomplish the performance simulation to compare and analyze the performance of each arbitration scheme according to the characteristics of the master and slave. With the performance simulation, we observed that when there are few masters on critical path in a bus system, the arbitration scheme based on dynamic priority shows the maximum performance and in other cases, the arbitration scheme based on round robin shows the highest performance. In addition, the arbitration scheme with transaction based multiplexing shows higher performance than the same arbitration scheme with single transfer based switching in an application with frequent accesses to the long latency devices or memories such as SDRAM. The improvements of the arbitration scheme with transaction based multiplexing are 26%, 42% and 51%, respectively when the latency times of SDRAM are 1, 2 and 3 clock cycles.