Daechul Park

A stable feedback control of the buffer state using the controlled Lagrange multiplier method

We propose a stable feedback control algorithm of the buffer state using the controlled Lagrange multiplier in combined rate-distortion characteristics for a buffer-constrained adaptive quantization. The proposed algorithm is established using feedback control theory and is described by the state equation with nonlinearity in the feedback path. The nonlinearity of the state equation comes from the average distortion-rate curve of the information source. The stability of this buffer control algorithm is shown using Lyapunov stability theory. It is shown that the stability of the algorithm depends on average distortion-rate curve and buffer size, and sufficient conditions for the algorithm to be stable are obtained from the usual average distortion-rate curve. In addition, it is observed from experimental results that the performance of the proposed algorithm and the optimal algorithm is not significantly different. Although the proposed algorithm does not provide optimal performance, the algorithm can be implemented easily with quite low computational complexity, as compared with others.

Image compression based on best wavelet packet bases

An adaptive subband decomposition technique for efficient signal representation and compression is introduced and tested in a rate-distortion framework. The best wavelet packet bases exhibit a SFFT (short-time fast Fourier transform) subband decomposition at one source instance, a wavelet decomposition at another instance, or any intermediate wavelet packet decomposition at yet other instances to best match the signals characteristics. For a given subband decomposition, commonly used information measures such as entropy, distortion (mse), and rate (no. of coefficients) are minimized over all subbands decomposition to search the most efficient wavelet packet tree of signal and a method to provide such a wavepacket tree is proposed. Image coding application results using the joint rate-distortion cost measure demonstrated superior performance over the entropy only information cost measure. >

Buffered rate-distortion control of MPEG compressed video channel for DBS applications

A joint rate-distortion control algorithm of multichannel sources is proposed to control the overall picture quality by putting some constraints on the encoders quantization parameters and buffer state to allow a limited variable bit rate (VBR) coding within a group of pictures (GOP) block. A major drawback for video transmission over a conventional fixed rate channel without using a VBR encoder is that received picture quality varies greatly, resulting in non-uniform subjective perception. In order to minimize these effects buffered-constrained rate-distortion control within a GOP block level finds an optimal solution in a combined rate-distortion sense by permitting a VBR encoding as well as the direct feedback of the buffer state into the cost measure when multichannel programs are statistically multiplexed. Results indicate some quality gains over the conventional constant bit rate method.

Adaptive Bayesian decision model for motion segmentation

Abstract A new method is proposed for modeling of the target and background event based on the classical decision theory using a Bayesian model. The implementation of an adaptive video motion detection system at the existing board level is also addressed using DATACUBE image processing hardware.

Statistical analysis of a change detector based on image modeling of difference picture

A statistical analysis of a change detector for motion detection based on image modeling of a difference picture is presented. The approach is founded upon the fact that complete silhouettes of moving objects can be segmented at each frame time using a change detector based on image modeling of the difference picture. The change detection problem can be treated as a signal detection problem in which very little is known about the signal to be detected. The exact modeling of the background event is necessary for a robust change detector which can adapt to changing environments. The governing statistics for change measures, which have a form of sum of squares of the difference picture, can be inferred from the assumed Gaussian distribution of the stationary background in the difference picture and are used for calculation of the adaptive decision threshold for the change detector. It is shown that the adaptive decision threshold derived from the change measure minimizes the total probability of error as measured in false alarms and missed detections. Experimental results that support the underlying statistical analysis are presented.<<ETX>>

Performance improvement of turbo codes by error range search method in Rayleigh fading channel

We propose an efficient method of decoding turbo codes that can greatly reduce the iteration number with very low-complexity. The proposed ERS method reduces the number of iteration to obtain the required BER (bit error rate). Consequently, the number of iterations and latency required by the iterative decoding operation in turbo codes can be reduced and the interleavers size can be minimized.

Image Features Based Secure Access Control for Data Content Protection

Data security is always an important issue. In particular, the current emerging cloud computing system inevitably raises the issue of data security. However, data security is no longer safe with a simple way, but requires rather advanced method to secure the data. In this paper, instead of exploiting the existing text-based cryptography approach an image-based access control of data content is studied to present a higher level of data security. Color key chain is generated both using histogram value of the original image, and the location information and featured color information extracted by geometric transformation to form the security key to access secure data content. Finally, the paper addresses design interface and implementation for data content access control for evaluation of the proposed scheme.

A Multiclass Deep Convolutional Neural Network Classifier for Detection of Common Rice Plant Anomalies

This study examines the use of deep convolutional neural network in the classification of rice plants according to health status based on images of its leaves. A three-class classifier was implemented representing normal, unhealthy, and snail-infested plants via transfer learning from an AlexNet deep network. The network achieved an accuracy of 91.23%, using stochastic gradient descent with mini batch size of thirty (30) and initial learning rate of 0.0001. Six hundred (600) images of rice plants representing the classes were used in the training. The training and testing dataset-images were captured from rice fields around the district and validated by technicians in the field of agriculture.

Experiments on An Network Processor-based Intrusion Detection

To help network intrusion detection systems(NIDSs) keep up with the demands of today`s networks, that we the increasing network throughput and amount of attacks, a radical new approach in hardware and software system architecture is required. In this paper, we propose a Network Processor(NP) based In-Line mode NIDS that supports the packet payload inspection detecting the malicious behaviors, as well as the packet filtering and the traffic metering. In particular, we separate the filtering and metering functions from the deep packet inspection function using two-level searching scheme, thus the complicated and time-consuming operation of the deep packet inspection function does not hinder or flop the basic operations of the In-line mode system. From a proto-type NP-based NIDS implemented at a PC platform with an x86 processor running Linux, two Gigabit Ethernet ports, and 2.5Gbps Agere PayloadPlus(APP) NP solution, the experiment results show that our proposed scheme can reliably filter and meter the full traffic of two gigabit ports at the first level even though it can inspect the packet payload up to 320 Mbps in real-time at the second level, which can be compared to the performance of general-purpose processor based Inspection. However, the simulation results show that the deep packet searching is also possible up to 2Gbps in wire speed when we adopt 10Gbps APP solution.

Implementation and Performance Evaluation of High-Performance Intrusion Detection and Response System

Recently, the growth of information infrastructure is getting fatter and faster. At the same time, the security accidents are increasing together. We have problem that do not handle traffic because we have the Intrusion Detection Systems in low speed environment. In order to overcome this, we need effective security analysis techniques that ran Processed data of high-capacity because high speed network environment. In this paper we proposed the Gigabit Intrusion Detection System for coordinated security function such as intrusion detection, response on the high speed network. We suggested the detection mechanism in high speed network environment that have pattern matching function based packet header and based packet data that is proceeded in system kernel area, we are shown that this mechanism was excellent until maximum 20 times than existing system in traffic processing performance.