• Proceedings of the Korean Society of Computer Information Conference
• /
• 2024.01a
• /
• pp.55-58
• /
• 2024

최근 각 기업의 AI 면접시스템 도입이 증가하고 있으며, AI 면접에 대한 실효성 논란 또한 많은 상황이다. 본 논문에서는 AI 면접 과정에서 지원자를 평가하는 방식을 시각, 음성, 자연어처리 3영역에서 구현함으로써, 면접 지원자를 다방면으로 분석 방법론의 적절성에 대해 평가하고자 한다. 첫째, 시각적 측면에서, 면접 지원자의 감정을 인식하기 위해, 합성곱 신경망(CNN) 기법을 활용해, 지원자 얼굴에서 6가지 감정을 인식했으며, 지원자가 카메라를 응시하고 있는지를 시계열로 도출하였다. 이를 통해 지원자가 면접에 임하는 태도와 특히 얼굴에서 드러나는 감정을 분석하는 데 주력했다. 둘째, 시각적 효과만으로 면접자의 태도를 파악하는 데 한계가 있기 때문에, 지원자 음성을 주파수로 환산해 특성을 추출하고, Bidirectional LSTM을 활용해 훈련해 지원자 음성에 따른 6가지 감정을 추출했다. 셋째, 지원자의 발언 내용과 관련해 맥락적 의미를 파악해 지원자의 상태를 파악하기 위해, 음성을 STT(Speech-to-Text) 기법을 이용하여 텍스트로 변환하고, 사용 단어의 빈도를 분석하여 지원자의 언어 습관을 파악했다. 이와 함께, 지원자의 발언 내용에 대한 감정 분석을 위해 KoBERT 모델을 적용했으며, 지원자의 성격, 태도, 직무에 대한 이해도를 파악하기 위해 객관적인 평가지표를 제작하여 적용했다. 논문의 분석 결과 AI 면접의 다면적 평가시스템의 적절성과 관련해, 시각화 부분에서는 상당 부분 정확도가 객관적으로 입증되었다고 판단된다. 음성에서 감정분석 분야는 면접자가 제한된 시간에 모든 유형의 감정을 드러내지 않고, 또 유사한 톤의 말이 진행되다 보니 특정 감정을 나타내는 주파수가 다소 집중되는 현상이 나타났다. 마지막으로 자연어처리 영역은 면접자의 발언에서 나오는 말투, 특정 단어의 빈도수를 넘어, 전체적인 맥락과 느낌을 이해할 수 있는 자연어처리 분석모델의 필요성이 더욱 커졌음을 판단했다.

• The Journal of Korea Institute of Information, Electronics, and Communication Technology
• /
• v.17 no.1
• /
• pp.76-81
• /
• 2024

Recently, agricultural methods are changing from experience-based agriculture to data-based agriculture. Changes in agricultural production due to the 4th Industrial Revolution are largely occurring in three areas: smart sensing and monitoring, smart analysis and planning, and smart control. In order to realize open-field smart agriculture, information on the physical and chemical properties of soil is essential. Conventional physicochemical measurements are conducted in a laboratory after collecting samples, which consumes a lot of cost, labor, and time, so they are quickly measured in the field. Measurement technology that can do this is urgently needed. In addition, a soil analysis system that can be carried and moved by the measurer and used in Korea's rice fields, fields, and facility houses is needed. To solve this problem, our goal is to develop and commercialize software that can collect soil samples and analyze the information. In this study, basic soil composition measurement was conducted using soil composition measurement sensors consisting of hardness measurement and electrode sensors. Through future research, we plan to develop a system that applies soil sampling using a CCD camera, ultrasonic sensor, and sampler. Therefore, we implemented a sensor and soil analysis UI that can measure and analyze the soil condition in real time, such as hardness measurement display using a load cell and moisture, PH, and EC measurement display using conductivity.

• Journal of the Korean earth science society
• /
• v.45 no.2
• /
• pp.111-120
• /
• 2024

Cloud coverage is a key factor in determining whether to proceed with observations. In the past, human judgment played an important role in weather evaluation for observations. However, the development of remote and robotic observation has diminished the role of human judgment. Moreover, it is not easy to evaluate weather conditions automatically because of the diverse cloud shapes and their rapid movement. In this paper, we present the development of a cloud monitoring program by applying a machine learning-based Python module "cloudynight" on all-sky camera images obtained at Miryang Arirang Astronomical Observatory (MAAO). The machine learning model was built by training 39,996 subregions divided from 1,212 images with altitude/azimuth angles and extracting 16 feature spaces. For our training model, the F1-score from the validation samples was 0.97, indicating good performance in identifying clouds in the all-sky image. As a result, this program calculates "Cloudiness" as the ratio of the number of total subregions to the number of subregions predicted to be covered by clouds. In the robotic observation, we set a policy that allows the telescope system to halt the observation when the "Cloudiness" exceeds 0.6 during the last 30 minutes. Following this policy, we found that there were no improper halts in the telescope system due to incorrect program decisions. We expect that robotic observation with the 0.7 m telescope at MAAO can be successfully operated using the cloud monitoring program.

• Korean Security Journal
• /
• no.24
• /
• pp.53-66
• /
• 2010

This study aims to analyze kinematic variables regarding lateral break-fall quantitatively that can protect the body, prevent the injury and minimize the impact. To this end, three Hapkido, judo athletes of H University with experience of over 5 years were selected. Test was conducted through three dimensional image analysis by checking the time and order the subjects reach a mat. In this study, lateral break-fall was repeated five times and among them, the best movement was selected. The picture shot with high-speed camera was analyzed by using KWON3D ver. 3.1 program through three dimensional coordinate calculation based on DLT method and smoothing process of data. Study results were as follows. 1. With respect to time variables shown in lateral break-fall of A, B, C athletes, there is small difference in temporal variables and in the order the body reaches a mat. With respect to average value, hand is ($0.94{\pm}0.20$), elbow ($0.97{\pm}0.17$), hip ($0.97{\pm}0.18$), back ($0.98{\pm}0.18$), and shoulder ($1.04{\pm}0.16$). Time variable the body reaches a mat in lateral break-fall is in hand, elbow, hip, back and shoulder. 2. With respect to moving distance variables shown in lateral break-fall of A, B, C athletes, hand is ($34.33{\pm}34.59$), elbow ($52.00{\pm}26.06$), hip ($70.00{\pm}15.72$), back ($153.67{\pm}17.93$), and should ($130.67{\pm}29.02$). The fact that this study contributed to improving security martial arts technique and protecting the body by understanding the principle of lateral break-fall movement is of significance. In addition, the fact that this study provided systematic basic data for improving security martial arts technique is significant.

• Korean Journal of Applied Biomechanics
• /
• v.12 no.2
• /
• pp.51-63
• /
• 2002

This study sought to identify the kinematic characteristics at entrance to the straight course from the curvilinear course in the 200m-track game. For this purpose, this study was conducted for 4 sprinters by setting the 10m-section combined from the curvilenear track to the straight course and shooting them with the camcorder. It was set up to include all the sections of analysis by using the framework of the control point knowing the coordinate of the space and actual analysis was conducted on the motion showing the best records by conducting it for each subject five times. As a result, the following conclusion was drawn: It was found that the subjects showed the average stride of 4.5${\pm}$0.41 times at the 10-meter section and the required time of 1.42${\pm}$0.04sec. They showed the ratio average stride to height of 1.25${\pm}$0.20% and the average speed of 7.06${\pm}$0.19m/s. The displacement in the center of gravity of the human body at the section combined from the curvilinear course to the straight course was moving along the inward course of the curvilinear course, and the displacement of the leg located at the outward direction(right) was found to be larger than that of the leg located at the inward direction(left). In the speed of the left and right hand segments, it was found that the speed of the right hand located in the outward direction was faster than that of the left hand located at the inward, and it was found that the subjects progressed in the curvilinear course. The subjects showed the larger angle of the shoulder joint when the upper arm was located in the forward direction than when the it was located in the backward direction. In the curvilinear course, they showed the lower value of the lateral angle of the trunk when the right foot located at the outward direction left the ground than when the left foot located at the inward direction left the ground. And it was found that the lateral angle of the trunk became lower with approaching the straight course.

• The Korean Journal of Nuclear Medicine Technology
• /
• v.15 no.1
• /
• pp.94-100
• /
• 2011

Purpose: Basal/Acetazolamide-challenged brain perfusion SPECT is very useful to assess cerebral perfusion and vascular reserve. However, as there is a trade off between sensitivity and spatial resolution in the selection of collimator, the selection of optimal collimator is crucial. In this study, we examined three collimators to select optimal one for 1-day brain perfusion SPECT. Materials and Methods: Three collimators, low energy high resolution-parallel beam (LEHR-par), ultra resolution-fan beam (LEUR-fan) and super fine-fan beam (LESFR-fan), were tested for 1-day imaging using Triad XLT 9 (TRIONIX). The SPECT images of Hoffman 3D brain phantom filled with 99mTc of 170 MBq and a normal volunteer were acquired with a protocol of 50 kcts/frame and detector rotation of 3 degree. Filterd backprojection (FBP) reconstruction with Butterworth filter (cut off frequencies, 0.3 to 0.5) was performed. The quantitative and qualitative assessments for three collimators were performed. Results: The blind tests showed that LESFR-fan provided the best image quality for Hoffman brain phantom and the volunteer. However, images for all the collimator were evaluated as 'acceptable'. On the other hand, in order to meet the equivalent signal-to-noise ratio (SNR), total acquisition time or radioactivity dose for LESFR-fan must have been increased up to almost twice of that for LEUR-fan and LEHR-par. The volunteer test indicated that total acquisition time could be reduced approximately by 10 to 14 min in clinical practice using LEUR-fan and LEHR-par without significant loss on image quality, in comparison with LESFR-fan. Conclusion: Although LESFR-fan provides the best image quality, it requires significantly more acquisition time than LEUR-fan and LEHR-par to provide reasonable SNR. Since there is no significant clinical difference between three collimators, LEUR-fan and LEHR-par can be recommended as optimal collimators for 1-day brain perfusion imaging with respect to image quality and SNR.

• Korean Journal of Materials Research
• /
• v.11 no.5
• /
• pp.427-430
• /
• 2001

We prepared 10cm-diameter Si(100)/500 $\AA$-Si$_3$N$_4$/Si(100) wafer Pairs adopting 500 $\AA$ -thick Si$_3$N$_4$layer as insulating layer between single crystal Si wafers. Si3N, is superior to conventional SiO$_2$ in insulating. We premated a p-type(100) Si wafer and 500 $\AA$ -thick LPCVD Si$_3$N$_4$∥Si (100) wafer in a class 100 clean room. The cremated wafers are separated in two groups. One group is treated to have hydrophobic surface and the other to have hydrophilic. We employed a FLA(fast linear annealing) bonder to enhance the bond strength of cremated wafers at the scan velocity of 0.1mm/sec with varying the heat input at the range of 400~1125W. We measured bonded area using a infrared camera and bonding strength by the razor blade crack opening method. We used high resolution transmission electron microscopy(HRTEM) to probe cross sectional view of bonded wafers. The bonded area of two groups was about 75%. The bonding strength of samples which have hydrophobic surface increased with heat input up to 1577mJ/$m^2$ However, bonding strength of samples which have hydrophilic surface was above 2000mJ/$m^2$regardless of heat input. The HRTEM results showed that the hydrophilic samples have about 25 $\AA$ -thick SiO layer between Si and Si$_3$N$_4$/Si and that maybe lead to increase of bonding strength.

• The Korean Journal of Nuclear Medicine Technology
• /
• v.19 no.1
• /
• pp.57-61
• /
• 2015

Purpose We considered the correlation of Ejection Fraction (EF) which was analyzed by Multi Gated Blood Pool Scan (MUGA) and Echocardiography (ECHO) for the patients who were classified according to the condition of cardiac function. Materials and Methods We analyzed the patients (female 60) who were diagnosed with breast cancer and were examined by both MUGA and ECHO. The 30 patients (age: $58.27{\pm}13.48$) who were analyzed into less than 50% to 70% of EF were categorized as normal group and the other 30 patients (age: $53.70{\pm}8.45$) who were analyzed into less than 50% of EF were categorized as abnormal group. Statistical analysis with SPSS ver. 18 was applied. Results Each of the value of mean and standard deviation of normal group was $66.43{\pm}5.80$ (MUGA), $60.50{\pm}4.93$ (ECHO). There was a significant difference (p<0.001). Each of the value of mean and standard deviation of abnormal group was $41.93{\pm}7.58$ (MUGA), $41.70{\pm}11.49$ (ECHO). There was no significant difference (p>0.001). In the result, all 30 cases of normal group showed the same reading. 8 out of 30 cases in abnormal group showed inconsistency of the reading. Conclusion We could confirm the correlation of the EF in MUGA and ECHO statistically. There was difference between abnormal groups from the result of reading. If we are aware of the result according to the different cardiac function categorization, MUGA and ECHO can be used as even more accurate interchangeable test.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.21 no.12
• /
• pp.558-569
• /
• 2020

This study examined the frequency of a broiler's stay in each area as stock density using an ICT-based image analysis technique from the perspective of precision livestock farming (PLF) according to the increase in the domestic broiler farms to understand the normal behavior patterns of broilers by age. The broiler was used in the experimental box (3.3×2.7 m) in a poultry house in Gyeonggi province. The stock densities were 9.5 birds/㎡ (n=85) and 19 birds/㎡ (n=170), respectively, and the frequency of stay by feeding, water, and rest area was monitored using a top-view camera. The image data of three-colored-specific broilers identified as the stock density were acquired by age (12, 16, 22, 27, and 29 days) for six hours. In the collected image data, the object tracking technique was used to record the cumulative movement path by connecting approximately 640,000 frames at 30 fps to quantify the frequency of stay in each area. In each stock density, it was significant in the order of the rest area, feeding, and water area (p<0.001). In 9.5 birds/㎡, it was at 57.9, 24.2, and 17.9 %, and 73.2, 16.8, and 10 % in 19 birds/㎡. The frequency of a broiler's stay could be evaluated in each area as the stock density using an ICT-based image analysis technique that minimizes stress. This method is expected to be used to provide basic material for developing an ICT-based management system through real-time monitoring.

• The Korean Journal of Nuclear Medicine Technology
• /
• v.25 no.1
• /
• pp.15-20
• /
• 2021

Purpose This article studies the relationship between the length of a kidney measured by two scanning methods: Kidney Computed Tomography (CT) and ^99mTc-Dimercaptosuccinic acid (DMSA) renal scan. Kidney CT provides a better anatomic assessment, while ^99mTc-DMSA renal scan is superior in the kidney function test. Materials and Methods From January to December of 2019, two hundred patients who had Kidney CT and Tc^99m-DMSA renal scan were chosen for this study. SPSS17.0 was selected for statistical analysis. Results Due to the effect of the breathing and resolution of ^99mTc-DMSA renal scan, it showed the kidney's relatively longer length than the length of Kidney CT. For the same kidney, the length comparison among different brands' Gamma cameras was negligible. The length difference within the same age group did not show a noticeable discrepancy. However, there was a length difference between the radio technologists. Kidney CT and ^99mTc-DMSA renal scan indicated a strong positive correlation between the length of the left and right kidney. Conclusion It is necessary to establish a standardized measurement method for measuring kidney length using ^99mTc-DMSA renal scan. The kidney's functional changes and length changes are indications of Kidney diseases. Especially, pediatric patients tend to use ^99mTc-DMSA renal scan for assessing the kidney's shape and the function to avoid potential radiation exposure during the Kidney CT. Therefore, it is significant to provide not only the kidney's functional information but also an anatomic analysis, including the kidney's size, length, and location through the ^99mTc-DMSA renal scan.