• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.18 no.7
• /
• pp.1706-1725
• /
• 2024

The unique U-shaped structure of U-Net network makes it achieve good performance in image segmentation. This network is a lightweight network with a small number of parameters for small image segmentation datasets. However, when the medical image to be segmented contains a lot of detailed information, the segmentation results cannot fully meet the actual requirements. In order to achieve higher accuracy of medical image segmentation, a novel improved U-Net network architecture called multi-scale encoder-decoder U-Net+ (MEDU-Net+) is proposed in this paper. We design the GoogLeNet for achieving more information at the encoder of the proposed MEDU-Net+, and present the multi-scale feature extraction for fusing semantic information of different scales in the encoder and decoder. Meanwhile, we also introduce the layer-by-layer skip connection to connect the information of each layer, so that there is no need to encode the last layer and return the information. The proposed MEDU-Net+ divides the unknown depth network into each part of deconvolution layer to replace the direct connection of the encoder and decoder in U-Net. In addition, a new combined loss function is proposed to extract more edge information by combining the advantages of the generalized dice and the focal loss functions. Finally, we validate our proposed MEDU-Net+ MEDU-Net+ and other classic medical image segmentation networks on three medical image datasets. The experimental results show that our proposed MEDU-Net+ has prominent superior performance compared with other medical image segmentation networks.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.67 no.11
• /
• pp.1506-1511
• /
• 2018

This study presents a human sensibility evaluation method using neural network and multiple-template method on electroencephalogram(EEG). We used a multi-layer perceptron type neural network as the sensibility classifier using EEG signal. For our research objective, 10-channel EEG signals are collected from the healthy subjects. After the necessary preprocessing is performed on the acquired signals, the various EEG parameters are estimated and their discriminating performance is evaluated in terms of pattern classification capability. In our study, Linear Prediction(LP) coefficients are utilized as the feature parameters extracting the characteristics of EEG signal, and a multi-layer neural network is used for indicating the degree of human sensibility. Also, the estimation for human comfortableness is performed by varying temperature and humidity environment factors and our results showed that the proposed scheme achieved good performances for evaluation of human sensibility.

• Journal of Korea Multimedia Society
• /
• v.21 no.8
• /
• pp.858-863
• /
• 2018

In this paper, we propose and implement a multi-level convolutional neural network (CNN) algorithm to identify the sexually explicit and lewdness of various image files, and verify its effectiveness by using unauthorized image files generated in the actual military. The proposed algorithm increases the accuracy by applying the convolutional artificial neural network step by step to minimize classification error between similar categories. Experimental data have categorized 20,005 images in the real field into 6 authorization categories and 11 non-authorization categories. Experimental results show that the overall detection rate is 99.51% for the image files. In particular, the excellence of the proposed algorithm is verified through reducing the identification error rate between similar categories by 64.87% compared with the general CNN algorithm.

• Journal of Korea Multimedia Society
• /
• v.19 no.3
• /
• pp.573-584
• /
• 2016

Recent application range of sensor networks is becoming diverse. It means collected sensor data types are becoming diverse too. These sensor data have their own characteristics. Thus achieving energy efficiency, existing sensor network management policy consider their own characteristics. However, it is inefficient to apply the existing network management schemes for controlling such kind of data at the same time. Because, existing network management schemes considered one type of data only. Therefore, we propose a novel routing scheme that is able to efficient energy conservation through effective data controlling on multi-sensor application environment.

• Proceedings of the KIEE Conference
• /
• 2005.05a
• /
• pp.206-208
• /
• 2005

Wireless sensor network's value has increased greatly in recent years in the fields of Ubiquitous Computing that function as solution to reduce both the limitation and collision about RFID Technology. The research for wireless sensor network technology is proceeding with the research for various sensor nodes, powerful routing algorithms, securities for data transmission, and valid applications. This paper suggests that we make the new multi-hop routing algorithm using RSS in order to implement enhanced multi-hop routing algorithm. This paper should demonstrate that the routing algorithm using suggested RSS is superior to routing algorithm based on established BSDV(Destination Sequenced Distance Vector).

• Journal of the Korean Society for Aviation and Aeronautics
• /
• v.21 no.1
• /
• pp.15-22
• /
• 2013

In oceanic flight routes, HF radio and satellite data links have weather restrictions, long propagation delay and low data throughput. In this paper, we propose oceanic aeronautical communications scheme in the VHF band based on ad-hoc communication. The proposed scheme organizes autonomously a multi-hop network that is divided into multiple local network using aircraft to fly long-distance communication and supports a hybrid type of multiple access, which consists of random access and TDMA (Time Division Multiple Access) scheme. In addition, several algorithms to apply spatial reuse of transmission to multi-hop long range communication environments have been proposed. The proposed system proves performance improvement on delay time as an effective solution to communicate end-to-end on the oceanic flight routes and strengthens the reliability of oceanic aeronautical communication.

• ETRI Journal
• /
• v.32 no.4
• /
• pp.622-625
• /
• 2010

In this letter, we assess the scalability of a path computation flooding (PCF) approach to compute optimal end-to-end inter-domain paths in a path computation element-based multi-domain network. PCF yields a drastically reduced network blocking probability compared to a blind per-domain path computation but introduces significant network control overhead and path computation complexity. In view of this, we introduce and compare an alternative low overhead PCF (LoPCF) solution. From the obtained results, LoPCF leads to similar blocking probabilities to PCF while exhibiting around 50% path computation complexity and network control overhead reduction.

• Journal of Institute of Control, Robotics and Systems
• /
• v.16 no.2
• /
• pp.114-119
• /
• 2010

Network technology has been developed rapidly for digital service in these days. ZigBee, one of the IEEE 802.15.4 protocols, supporting local communication has become the core technology in the wireless network area. In this paper we designed an integrated fire monitoring system using a mobile robot and the ZigBee sensor nodes which are deployed to monitor fires. When a fire breaks out, the image information of the scene of a fire is transmitted by an autonomous mobile robot and we also monitor the current position of the robot. Furthermore, the data around the place where the fire breaks out and the positions of the sensor nodes can be transmitted to a server via the multi-hop communication in the real time.

• Proceedings of the KIEE Conference
• /
• 1996.11a
• /
• pp.62-64
• /
• 1996

Neural network as a controller of a nonlinear system and a system identifier has been studied during the past few years. A well trained neural network identifier can be used as a system predictor. We proposed the method to design multi-step ahead predictor and multi-step predictive controller using neural network. We used the input and out put data of B system to train the NNP and used the forecasted approximat system output from NNP as B input of NNC. In this paper we used two-step ahead predictive controller to test B heating controll system and compared with PI controller.

• Journal of information and communication convergence engineering
• /
• v.11 no.1
• /
• pp.12-16
• /
• 2013

Sensor network is one of the strongest technologies for various applications including home automation, surveillance system and monitoring system. To ensure secure and robust network communication between sensor nodes, plain-text should be encrypted using encryption methods. However due to their limited computation power and storage, it is difficult to implement public key cryptography, including elliptic curve cryptography, RSA and pairing cryptography, on sensor networks. However, recent works have shown the possibility that public key cryptography could be made available in a sensor network environment by introducing the efficient multi-precision multiplication method. The previous method suggested a broad rule of multiplication to enhance performance. However, various features of sensor motes have not been considered. For optimized implementation, unique features should be handled. In this paper, we propose a fully optimized multiplication method depending on a different specification for sensor motes. The method improves performance by using more efficient instructions and general purpose registers.