• Journal of KIISE:Information Networking
• /
• v.32 no.2
• /
• pp.147-155
• /
• 2005

For the last several decades, many researches have been performed to distribute bandwidth fairly between sessions. In this problem, the most important challenge is to realize a scalable implementation and high fairness simultaneously. Here high fairness means that bandwidth is distributed fairly even in short time intervals. Unfortunately, existing scheduling algorithms either are lack of scalable implementation or can achieve low fairness. In this paper, we propose a scheduling algorithm that can achieve feasible fairness without losing scalability. The proposed algorithm is a Hierarchical Deficit Round-Robin (H-DRR). While H-DRR requires a constant time for implementation, the achievable fairness is similar to that of Packet-by-Packet Generalized Processor Sharing(PGPS) algorithm. PGPS has worse scalability since it uses a sorted-priority queue requiring O(log N) implementation complexity where N is the number of sessions.

• Journal of KIISE:Information Networking
• /
• v.32 no.2
• /
• pp.254-260
• /
• 2005

In the guaranteed service, a real-time scheduling algorithm must achieve both high level of network utilization and scalable implementation. Here, network utilization indicates the number of admitted real-time sessions. Unfortunately, existing scheduling algorithms either are lack of scalable implementation or can achieve low network utilization. For example, scheduling algorithms based on time-stamps have the problem of O(log N) scheduling complexity where N is the number of sessions. On the contrary, round-robin algorithms require O(1) complexity. but can achieve just a low level of network utilization. In this paper, we propose a scheduling algorithm that can achieve high network utilization without losing scalability. The proposed algorithm is a Hierarchical Round-Robin (H-RR) algorithm that utilizes multiple rounds with different interval sizes. It provides latency bounds similar to those by Packet-by-Packet Generalized Processor Sharing (PGPS) algorithm using a sorted-Priority queue. However, H-RR requires a constant time for implementation.

• Journal of Intelligence and Information Systems
• /
• v.28 no.1
• /
• pp.89-106
• /
• 2022

Recently, the amount of video data collected from smartphones, CCTVs, black boxes, and high-definition cameras has increased rapidly. According to the increasing video data, the requirements for analysis and utilization are increasing. Due to the lack of skilled manpower to analyze videos in many industries, machine learning and artificial intelligence are actively used to assist manpower. In this situation, the demand for various computer vision technologies such as object detection and tracking, action detection, emotion detection, and Re-ID also increased rapidly. However, the object detection and tracking technology has many difficulties that degrade performance, such as re-appearance after the object's departure from the video recording location, and occlusion. Accordingly, action and emotion detection models based on object detection and tracking models also have difficulties in extracting data for each object. In addition, deep learning architectures consist of various models suffer from performance degradation due to bottlenects and lack of optimization. In this study, we propose an video analysis system consists of YOLOv5 based DeepSORT object tracking model, SlowFast based action recognition model, Torchreid based Re-ID model, and AWS Rekognition which is emotion recognition service. Proposed model uses single-linkage hierarchical clustering based Re-ID and some processing method which maximize hardware throughput. It has higher accuracy than the performance of the re-identification model using simple metrics, near real-time processing performance, and prevents tracking failure due to object departure and re-emergence, occlusion, etc. By continuously linking the action and facial emotion detection results of each object to the same object, it is possible to efficiently analyze videos. The re-identification model extracts a feature vector from the bounding box of object image detected by the object tracking model for each frame, and applies the single-linkage hierarchical clustering from the past frame using the extracted feature vectors to identify the same object that failed to track. Through the above process, it is possible to re-track the same object that has failed to tracking in the case of re-appearance or occlusion after leaving the video location. As a result, action and facial emotion detection results of the newly recognized object due to the tracking fails can be linked to those of the object that appeared in the past. On the other hand, as a way to improve processing performance, we introduce Bounding Box Queue by Object and Feature Queue method that can reduce RAM memory requirements while maximizing GPU memory throughput. Also we introduce the IoF(Intersection over Face) algorithm that allows facial emotion recognized through AWS Rekognition to be linked with object tracking information. The academic significance of this study is that the two-stage re-identification model can have real-time performance even in a high-cost environment that performs action and facial emotion detection according to processing techniques without reducing the accuracy by using simple metrics to achieve real-time performance. The practical implication of this study is that in various industrial fields that require action and facial emotion detection but have many difficulties due to the fails in object tracking can analyze videos effectively through proposed model. Proposed model which has high accuracy of retrace and processing performance can be used in various fields such as intelligent monitoring, observation services and behavioral or psychological analysis services where the integration of tracking information and extracted metadata creates greate industrial and business value. In the future, in order to measure the object tracking performance more precisely, there is a need to conduct an experiment using the MOT Challenge dataset, which is data used by many international conferences. We will investigate the problem that the IoF algorithm cannot solve to develop an additional complementary algorithm. In addition, we plan to conduct additional research to apply this model to various fields' dataset related to intelligent video analysis.

• The Journal of the Korea Contents Association
• /
• v.8 no.1
• /
• pp.169-176
• /
• 2008

In this paper, we develop radio resource management queueing model in IEEE 802.IS networks considering both connection and packet level. In the upper level connection, we model connection admission control depending on availability of bandwidth and priority queue in each service class. In the lower level packet, we model dynamic bandwidth allocation considering threshold and availability of bandwidth in each service class simultaneously. Hierarchical model is built using an extended Petri Nets, SRN (Stochastic Reward Nets). Bandwidth utilization and normal throughput as performance index for all service classes of traffic are calculated and numerical results are obtained.

• Journal of Communications and Networks
• /
• v.18 no.5
• /
• pp.784-795
• /
• 2016

In this paper, we investigate the resource allocation problem in time-varying heterogeneous wireless networks (HetNet) with multi-homing user equipments (UE). The stochastic optimization model is employed to maximize the network utility, which is defined as the difference between the HetNet's throughput and the total energy consumption cost. In harmony with the hierarchical architecture of HetNet, the problem of stochastic optimization of resource allocation is decomposed into two subproblems by the Lyapunov optimization theory, associated with the flow control in transport layer and the power allocation in physical (PHY) layer, respectively. For avoiding the signaling overhead, outdated dynamic information, and scalability issues, the distributed resource allocation method is developed for solving the two subproblems based on the primal-dual decomposition theory. After that, the adaptive resource allocation algorithm is developed to accommodate the timevarying wireless network only according to the current network state information, i.e. the queue state information (QSI) at radio access networks (RAN) and the channel state information (CSI) of RANs-UE links. The tradeoff between network utility and delay is derived, where the increase of delay is approximately linear in V and the increase of network utility is at the speed of 1/V with a control parameter V. Extensive simulations are presented to show the effectiveness of our proposed scheme.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.45 no.1
• /
• pp.47-54
• /
• 2008

In this paper, we propose an auto-detection algorithm of basal cell carcinoma(BCC) from the protophorphyrin IX(PpIX) fluorescence image induced by appling the methyl 5-aminolaevulinate(MAL) ointment-induced protophorphyrin IX(PpIX) to the skin tumour area and then shining the wood lamp on the area. The proposed algorithm first generates 3 mask areas-tumor area, suspected tumor area and tumor free area and then applies local watershed algorithm to the turner and the suspected tumor areas to make small watershed regions that include similar luminance value pixels. Next, small watershed regions are merged by hierarchical queue based fast region merging that uses the difference between the average luminance values of adjacent watershed regions as a region merging criterion and finally BCC regions are detected. 50 tissue samples are acquired from the tumour regions of 10 patients with BCC that are extracted by using the proposed algorithm and are performed pathological examination by expert dermatologist. Experiment result shows the rate of tumor detection from BCC lesion using presurgical in vivo of MAL-indeuced PpIX fluorescence has high sensitivity 94.1% and relatively high specificity 82.6%.