• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.38C no.7
• /
• pp.623-629
• /
• 2013

This paper presents the implementation result of the EEG(electroencephalogram) signal transmission protocol and its test platform. EEG measured by a dry-type electrode is directly converted into digital signal by ADC(analog-to-digital converter). Thereafter it is transferred DSP(digital signal processor) platform by $I^2C$(inter-integrated circuit) protocol. DSP conducts the pre-processing of EEG and extracts feature vectors of EEG. In this work, we implement the $I^2C$ protocol with 16 channels by using 10 or 12-bit ADC. In the implementation results, the overhead ratio for the 4 bytes data burst transmission measures 2.16 and the total data rates are 345.6 kbps and 414.72 kbps with 10-bit and 12-bit 1 ksps ADC, respectively. Therefore, in order to support a high speed mode of $I^2C$ for 400 kbps, it is required to use 16:1 and $(8:1){\times}2$ ratios for slave:master in 10-bit ADC and 12-bit ADC, respectively.

• The Journal of Korean Institute of Electromagnetic Engineering and Science
• /
• v.13 no.2
• /
• pp.176-187
• /
• 2002

The digital Audio Broadcasting (DAB) system adopts Coded OFDM(COFDM) for channel coding. The COFDM is a combined technique of multicarrier transmission(OFDM) and punctured convolutional coding with viterbi error correction. Because the channel coding is an important topic for OFDM systems, this paper proposes a new turbo coded OFDM system that replaces the existing RCPC codec by a turbo codec without modifying the puncturing procedure and puncturing vectors defined in the standard DAB system for compatibility. The performance of a new system is compared to that of the conventional system under the frequency selective Rician fading channel and the frequency selective Rayleigh fading channel in conjunction with DAB transmission mode I suitable for the terrestrial single frequency network(SFN) broadcasting. The standard system's performance was improved with the aid of turbo codec.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.15 no.2
• /
• pp.355-363
• /
• 2011

Recently, as one of several methods to improve the performance of DVC(Distributed Video Coding) system, many research works are focusing on the iterative refinement of side information. Most of the conventional techniques are mainly based on the relationship between the reconstruction level and side information, or the vector median filtering of motion vectors, but, their performance improvements are restricted. In order to overcome the performance limit of the conventional schemes, in this paper, a side information generation scheme is designed by measuring the block-cost estimation. Then, by adaptively selecting the compensation mode using the received bit-plane information, we propose a block-adaptive iterative refinement which is efficient for non-symmetric moving objects. Computer simulations show that, by using the proposed refinement method, the performance can be improved up to 0.2 dB in rate-distortion.

• The KIPS Transactions:PartB
• /
• v.15B no.3
• /
• pp.205-210
• /
• 2008

A new approach is proposed for restoration of corrupted regions and segmentation in natural text images. The challenge is to fill in the corrupted regions on the basis of color feature analysis by second order symmetric stick tensor. It is show how feature analysis can benefit from analyzing features using tensor voting with chromatic and achromatic components. The proposed method is applied to text images corrupted by manifold types of various noises. Firstly, we decompose an image into chromatic and achromatic components to analyze images. Secondly, selected feature vectors are analyzed by second-order symmetric stick tensor. And tensors are redefined by voting information with neighbor voters, while restore the corrupted regions. Lastly, mode estimation and segmentation are performed by adaptive mean shift and separated clustering method respectively. This approach is automatically done, thereby allowing to easily fill-in corrupted regions containing completely different structures and surrounding backgrounds. Applications of proposed method include the restoration of damaged text images; removal of superimposed noises or streaks. We so can see that proposed approach is efficient and robust in terms of restoring and segmenting text images corrupted.

• Journal of the Institute of Electronics Engineers of Korea SC
• /
• v.44 no.5
• /
• pp.62-70
• /
• 2007

In this paper, the boost Power Factor Correction(PFC) technique for Direct Torque Control(DTC) of brushless DC motor drive in the constant torque region is implemented on a TMS320F2812DSP. Unlike conventional six-step PWM current control, by properly selecting the inverter voltage space vectors of the two-phase conduction mode from a simple look-up table at a predefined sampling time, the desired quasi-square wave current is obtained, therefore a much faster torque response is achieved compared to conventional current control. Furthermore, to eliminate the low-frequency torque oscillations caused by the non-ideal trapezoidal shape of the actual back-EMF waveform of the BLDC motor, a pre-stored back-EMF versus position look-up table is designed. The duty cycle of the boost converter is determined by a control algorithm based on the input voltage, output voltage which is the dc-link of the BLDC motor drive, and inductor current using average current control method with input voltage feed-forward compensation during each sampling period of the drive system. With the emergence of high-speed digital signal processors(DSPs), both PFC and simple DTC algorithms can be executed during a single sampling period of the BLDC motor drive. In the proposed method, since no PWM algorithm is required for DTC or BLDC motor drive, only one PWM output for the boost converter with 80 kHz switching frequency is used in a TMS320F2812 DSP. The validity and effectiveness of the proposed DTC of BLDC motor drive scheme with PFC are verified through the experimental results. The test results verify that the proposed PFC for DTC of BLDC motor drive improves power factor considerably from 0.77 to as close as 0.9997 with and without load conditions.

• Journal of Life Science
• /
• v.22 no.8
• /
• pp.1018-1023
• /
• 2012

Snail is a zinc finger transcription factor that induces epithelial-to-mesenchymal transition (EMT), which promotes tumor invasion and metastasis by repressing E-cadherin expression. In addition, Snail restricts the cellular apoptotic response to apoptotic stimuli or survival factor withdrawal; however, its molecular mechanism remains largely unknown. In this study, we have investigated the mechanism underlying Snail-mediated chemoresistance to 5-fluorouracil (5-FU), one of the most widely used anti-cancer drugs. When Snail was overexpressed by doxycycline (DOX) in MCF-7 #5 cells, it inhibited 5-FU-induced apoptotic cell death and switched the cell death mode to necrosis. Snail expression, either by DOX treatment in MCF-7 #5 cells or by the transfection of Snail expression vectors pCR3.1-Snail-Flg, phosphorylation-resistant pCR3.1-S104, and 107A Snail-Flg in MCF-7 cells specifically induced PTEN down-regulation/inactivation and Akt/PKB activation, without affecting ERK1/2 activity. In addition, Snail prominently suppressed 5-FU-induced increases in p53 levels. These findings demonstrate that Snail switches 5-FU-induced apoptosis to necrosis through the activation of Akt/PKB and the down-regulation of p53 levels.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.49 no.2
• /
• pp.149-156
• /
• 2012

Support vector machines (SVMs) are well known for their pattern recognition capability, but proper care should be taken to alleviate their inherent implementation cost resulting from high computational intensity and memory requirement, especially in embedded systems where only limited resources are available. Since the memory requirement determined by the dimensionality and the number of support vectors is generally too high for a cache in embedded systems to accomodate, frequent accesses to the main memory occur inevitably whenever the cache is not able to provide requested data to the processor. These frequent accesses to the main memory result in overall performance degradation and increased energy consumption because a memory access typically takes longer and consumes more energy than a cache access or a register access. In this paper, we propose a technique that reduces the number of main memory accesses by optimizing the data access pattern of the SVM-based classifier in such a way that the temporal locality of the accesses increases, fully utilizing data loaded into the processor chip. With experiments, we confirm the enhancement made by the proposed technique in terms of the number of memory accesses, overall execution time, and energy consumption.

• Journal of Intelligence and Information Systems
• /
• v.26 no.2
• /
• pp.1-25
• /
• 2020

In this paper, we suggest an application system architecture which provides accurate, fast and efficient automatic gasometer reading function. The system captures gasometer image using mobile device camera, transmits the image to a cloud server on top of private LTE network, and analyzes the image to extract character information of device ID and gas usage amount by selective optical character recognition based on deep learning technology. In general, there are many types of character in an image and optical character recognition technology extracts all character information in an image. But some applications need to ignore non-of-interest types of character and only have to focus on some specific types of characters. For an example of the application, automatic gasometer reading system only need to extract device ID and gas usage amount character information from gasometer images to send bill to users. Non-of-interest character strings, such as device type, manufacturer, manufacturing date, specification and etc., are not valuable information to the application. Thus, the application have to analyze point of interest region and specific types of characters to extract valuable information only. We adopted CNN (Convolutional Neural Network) based object detection and CRNN (Convolutional Recurrent Neural Network) technology for selective optical character recognition which only analyze point of interest region for selective character information extraction. We build up 3 neural networks for the application system. The first is a convolutional neural network which detects point of interest region of gas usage amount and device ID information character strings, the second is another convolutional neural network which transforms spatial information of point of interest region to spatial sequential feature vectors, and the third is bi-directional long short term memory network which converts spatial sequential information to character strings using time-series analysis mapping from feature vectors to character strings. In this research, point of interest character strings are device ID and gas usage amount. Device ID consists of 12 arabic character strings and gas usage amount consists of 4 ~ 5 arabic character strings. All system components are implemented in Amazon Web Service Cloud with Intel Zeon E5-2686 v4 CPU and NVidia TESLA V100 GPU. The system architecture adopts master-lave processing structure for efficient and fast parallel processing coping with about 700,000 requests per day. Mobile device captures gasometer image and transmits to master process in AWS cloud. Master process runs on Intel Zeon CPU and pushes reading request from mobile device to an input queue with FIFO (First In First Out) structure. Slave process consists of 3 types of deep neural networks which conduct character recognition process and runs on NVidia GPU module. Slave process is always polling the input queue to get recognition request. If there are some requests from master process in the input queue, slave process converts the image in the input queue to device ID character string, gas usage amount character string and position information of the strings, returns the information to output queue, and switch to idle mode to poll the input queue. Master process gets final information form the output queue and delivers the information to the mobile device. We used total 27,120 gasometer images for training, validation and testing of 3 types of deep neural network. 22,985 images were used for training and validation, 4,135 images were used for testing. We randomly splitted 22,985 images with 8:2 ratio for training and validation respectively for each training epoch. 4,135 test image were categorized into 5 types (Normal, noise, reflex, scale and slant). Normal data is clean image data, noise means image with noise signal, relfex means image with light reflection in gasometer region, scale means images with small object size due to long-distance capturing and slant means images which is not horizontally flat. Final character string recognition accuracies for device ID and gas usage amount of normal data are 0.960 and 0.864 respectively.