• The Journal of the Acoustical Society of Korea
• /
• v.26 no.6
• /
• pp.250-258
• /
• 2007

Current speech recognition technology s achieved high performance with the development of hardware devices, however it is insufficient for some applications where high reliability is required, such as voice control of powered wheelchairs for disabled persons. For the system which aims to operate powered wheelchairs safely by voice in real environment, we need to consider that non-voice commands such as user s coughing, breathing, and spark-like mechanical noise should be rejected and the wheelchair system need to recognize the speech commands affected by disability, which contains specific pronunciation speed and frequency. In this paper, we propose non-voice rejection method to perform voice/non-voice classification using both YIN based fundamental frequency(F0) extraction and reliability in preprocessing. We adopted a multi-template dictionary and acoustic modeling based speaker adaptation to cope with the pronunciation variation of inarticulately uttered speech. From the recognition tests conducted with the data collected in real environment, proposed YIN based fundamental extraction showed recall-precision rate of 95.1% better than that of 62% by cepstrum based method. Recognition test by a new system applied with multi-template dictionary and MAP adaptation also showed much higher accuracy of 99.5% than that of 78.6% by baseline system.

• Journal of Practical Engineering Education
• /
• v.16 no.3_spc
• /
• pp.391-399
• /
• 2024

All highway toll gates in Korea use low-speed WIM(Weight-In-Motion) to block overloaded cargo vehicles from entering the main highway, but some cargo vehicle owners are illegally modifying vehicles to operate variable axles and evading crackdowns by manipulating the axles. In previous studies detect all tires of a running vehicle were detected to determine whether there is axle manipulation. However, because the vehicle entry area at the highway toll gate checkpoint is very narrow, there is a problem that it is realistically difficult to film all tires of the entering vehicle in one video frame. In this paper, we proposed a system that can determine whether the axle is being operated through changes in the height of the vehicle's cargo box rather than by detecting tires. To detect changes in the height of a cargo box, we propose a method to extract the representative line of the cargo box using Hough transform and then measure the change in height of the representative line to detect the change in height of the cargo box. In addition, we propose a method to detect changes in the vertical height of a cargo box by accumulating motion vectors of pixels within a certain area of the image using optical flow. And the two methods were compared and their advantages and disadvantages were analyzed and presented.

• The Journal of Korea Institute of Information, Electronics, and Communication Technology
• /
• v.17 no.4
• /
• pp.205-212
• /
• 2024

This study developed a deep learning model that distinguishes the front (with garnish) and the back (without garnish) surface of the dried semi-finished product (dried bukak) for screening operation before transfter the dried bukak to oil heater using robot's vacuum gripper. For deep learning model training and verification, RGB images for the front and back surfaces of 400 dry bukak that treated by data preproccessing were obtained. YOLO-v5 was used as a base structure of deep learning model. The area, surface information labeling, and data augmentation techniques were applied from the acquired image. Parameters including mAP, mIoU, accumulation, recall, decision, and F1-score were selected to evaluate the performance of the developed YOLO-v5 deep learning model-based surface detection model. The mAP and mIoU on the front surface were 0.98 and 0.96, respectively, and on the back surface, they were 1.00 and 0.95, respectively. The results of binary classification for the two front and back classes were average 98.5%, recall 98.3%, decision 98.6%, and F1-score 98.4%. As a result, the developed model can classify the surface information of the dried bukak using RGB images, and it can be used to develop a robot-automated system for the surface detection process of the dried bukak before deep frying.

• Journal of Intelligence and Information Systems
• /
• v.19 no.3
• /
• pp.113-125
• /
• 2013

Every company wants to know customer's requirement and makes an effort to meet them. Cause that, communication between customer and company became core competition of business and that important is increasing continuously. There are several strategies to find customer's needs, but VOC (Voice of customer) is one of most powerful communication tools and VOC gathering by several channels as telephone, post, e-mail, website and so on is so meaningful. So, almost company is gathering VOC and operating VOC system. VOC is important not only to business organization but also public organization such as government, education institute, and medical center that should drive up public service quality and customer satisfaction. Accordingly, they make a VOC gathering and analyzing System and then use for making a new product and service, and upgrade. In recent years, innovations in internet and ICT have made diverse channels such as SNS, mobile, website and call-center to collect VOC data. Although a lot of VOC data is collected through diverse channel, the proper utilization is still difficult. It is because the VOC data is made of very emotional contents by voice or text of informal style and the volume of the VOC data are so big. These unstructured big data make a difficult to store and analyze for use by human. So that, the organization need to automatic collecting, storing, classifying and analyzing system for unstructured big VOC data. This study propose an intelligent VOC analyzing system based on opinion mining to classify the unstructured VOC data automatically and determine the polarity as well as the type of VOC. And then, the basis of the VOC opinion analyzing system, called domain-oriented sentiment dictionary is created and corresponding stages are presented in detail. The experiment is conducted with 4,300 VOC data collected from a medical website to measure the effectiveness of the proposed system and utilized them to develop the sensitive data dictionary by determining the special sentiment vocabulary and their polarity value in a medical domain. Through the experiment, it comes out that positive terms such as "칭찬, 친절함, 감사, 무사히, 잘해, 감동, 미소" have high positive opinion value, and negative terms such as "퉁명, 뭡니까, 말하더군요, 무시하는" have strong negative opinion. These terms are in general use and the experiment result seems to be a high probability of opinion polarity. Furthermore, the accuracy of proposed VOC classification model has been compared and the highest classification accuracy of 77.8% is conformed at threshold with -0.50 of opinion classification of VOC. Through the proposed intelligent VOC analyzing system, the real time opinion classification and response priority of VOC can be predicted. Ultimately the positive effectiveness is expected to catch the customer complains at early stage and deal with it quickly with the lower number of staff to operate the VOC system. It can be made available human resource and time of customer service part. Above all, this study is new try to automatic analyzing the unstructured VOC data using opinion mining, and shows that the system could be used as variable to classify the positive or negative polarity of VOC opinion. It is expected to suggest practical framework of the VOC analysis to diverse use and the model can be used as real VOC analyzing system if it is implemented as system. Despite experiment results and expectation, this study has several limits. First of all, the sample data is only collected from a hospital web-site. It means that the sentimental dictionary made by sample data can be lean too much towards on that hospital and web-site. Therefore, next research has to take several channels such as call-center and SNS, and other domain like government, financial company, and education institute.

• Journal of the Korean Society of Fisheries and Ocean Technology
• /
• v.33 no.3
• /
• pp.159-165
• /
• 1997

This paper presents the optimum dimension of 89 ton class stow-net vessel with stern-fishing. The model of basic design is developed by using the optimization techniques referring to objective function and numerous constraints as follows; speed, fishing quantity, fishing days, catch per unit effort(CPUE), and weight/ratio of main dimensions, etc. Thus, the basic design of stow-net fishing vessel is built up by using the optimization of the design variables called the economic optimization criteria, and the objective function represents the criterion which is cost benefit ratio(CBR). The main conclusions are as follows. 1. S/W for decision of optimum hull size is developed in 89 ton class stow-net fishing vessel which is constructed with optimization of the design variables called the economic optimization criteria. 2. For optimum ship dimensions in 89 ton class stow-net fishing vessel, the hull dimensions can be obtained in the range of L= 27.3m, B = 6.6m, D = 2.80m, Cb = 0.695, T/D = 0.80, $\Delta$(displacement)=281.7ton with 10 knots.

• Journal of Intelligence and Information Systems
• /
• v.19 no.2
• /
• pp.125-140
• /
• 2013

We call a data set in which the number of records belonging to a certain class far outnumbers the number of records belonging to the other class, 'imbalanced data set'. Most of the classification techniques perform poorly on imbalanced data sets. When we evaluate the performance of a certain classification technique, we need to measure not only 'accuracy' but also 'sensitivity' and 'specificity'. In a customer churn prediction problem, 'retention' records account for the majority class, and 'churn' records account for the minority class. Sensitivity measures the proportion of actual retentions which are correctly identified as such. Specificity measures the proportion of churns which are correctly identified as such. The poor performance of the classification techniques on imbalanced data sets is due to the low value of specificity. Many previous researches on imbalanced data sets employed 'oversampling' technique where members of the minority class are sampled more than those of the majority class in order to make a relatively balanced data set. When a classification model is constructed using this oversampled balanced data set, specificity can be improved but sensitivity will be decreased. In this research, we developed a hybrid model of support vector machine (SVM), artificial neural network (ANN) and decision tree, that improves specificity while maintaining sensitivity. We named this hybrid model 'hybrid SVM model.' The process of construction and prediction of our hybrid SVM model is as follows. By oversampling from the original imbalanced data set, a balanced data set is prepared. SVM_I model and ANN_I model are constructed using the imbalanced data set, and SVM_B model is constructed using the balanced data set. SVM_I model is superior in sensitivity and SVM_B model is superior in specificity. For a record on which both SVM_I model and SVM_B model make the same prediction, that prediction becomes the final solution. If they make different prediction, the final solution is determined by the discrimination rules obtained by ANN and decision tree. For a record on which SVM_I model and SVM_B model make different predictions, a decision tree model is constructed using ANN_I output value as input and actual retention or churn as target. We obtained the following two discrimination rules: 'IF ANN_I output value <0.285, THEN Final Solution = Retention' and 'IF ANN_I output value ${\geq}0.285$, THEN Final Solution = Churn.' The threshold 0.285 is the value optimized for the data used in this research. The result we present in this research is the structure or framework of our hybrid SVM model, not a specific threshold value such as 0.285. Therefore, the threshold value in the above discrimination rules can be changed to any value depending on the data. In order to evaluate the performance of our hybrid SVM model, we used the 'churn data set' in UCI Machine Learning Repository, that consists of 85% retention customers and 15% churn customers. Accuracy of the hybrid SVM model is 91.08% that is better than that of SVM_I model or SVM_B model. The points worth noticing here are its sensitivity, 95.02%, and specificity, 69.24%. The sensitivity of SVM_I model is 94.65%, and the specificity of SVM_B model is 67.00%. Therefore the hybrid SVM model developed in this research improves the specificity of SVM_B model while maintaining the sensitivity of SVM_I model.

• The Korean Journal of Nuclear Medicine
• /
• v.35 no.3
• /
• pp.131-141
• /
• 2001

Purpose: We evaluated the sensitivity of the F-18 FDG PET by visual assessment and statistical parametric mapping (SPM) analysis for the localization of the epileptogenic zones in frontal lobe epilepsy. Materials and Methods: Twenty-four patients with frontal lobe epilepsy were examined. All patients exhibited improvements after surgical resection (Engel class I or II). Upon pathological examination, 18 patients revealed cortical dysplasia, 4 patients revealed tumor, and 2 patients revealed cortical scar. The hypometabolic lesions were found in F-18 FDG PET by visual assessment and SPM analysis. On SPM analysis, cutoff threshold was changed. Results: MRI showed structural lesions in 12 patients and normal results in the remaining 12. F-18 FDG PET correctly localized epileptogenic zones in 13 patients (54%) by visual assessment. Sensitivity of F-18 FDG PET in MR-negative patients (50%) was similar to that in MR-positive patients (67%). On SPM analysis, sensitivity decreased according to the decrease of p value. Using uncorrected p value of 0.05 as threshold, sensitivity of SPM analysis was 53%, which was not statistically different from that of visual assessment. Conclusion: F-18 FDG PET was sensitive in finding epileptogenic zones by revealing hypometabolic areas even in MR-negative patients with frontal lobe epilepsy as well as in MR-positive patients. SPM analysis showed comparable sensitivity to visual assessment and could be used as an aid in the diagnosis of epileptogenic zones in frontal lobe epilepsy.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.40 no.5
• /
• pp.312-321
• /
• 2003

In this paper, we implement robot which are ability to recognize obstacles and moving automatically to destination. we present two results in this paper; hardware implementation of image processing board and software implementation of visual feedback algorithm for a self-controlled robot. In the first part, the mobile robot depends on commands from a control board which is doing image processing part. We have studied the self controlled mobile robot system equipped with a CCD camera for a long time. This robot system consists of a image processing board implemented with DSPs, a stepping motor, a CCD camera. We will propose an algorithm in which commands are delivered for the robot to move in the planned path. The distance that the robot is supposed to move is calculated on the basis of the absolute coordinate and the coordinate of the target spot. And the image signal acquired by the CCD camera mounted on the robot is captured at every sampling time in order for the robot to automatically avoid the obstacle and finally to reach the destination. The image processing board consists of DSP (TMS320VC33), ADV611, SAA7111, ADV7l76A, CPLD(EPM7256ATC144), and SRAM memories. In the second part, the visual feedback control has two types of vision algorithms: obstacle avoidance and path planning. The first algorithm is cell, part of the image divided by blob analysis. We will do image preprocessing to improve the input image. This image preprocessing consists of filtering, edge detection, NOR converting, and threshold-ing. This major image processing includes labeling, segmentation, and pixel density calculation. In the second algorithm, after an image frame went through preprocessing (edge detection, converting, thresholding), the histogram is measured vertically (the y-axis direction). Then, the binary histogram of the image shows waveforms with only black and white variations. Here we use the fact that since obstacles appear as sectional diagrams as if they were walls, there is no variation in the histogram. The intensities of the line histogram are measured as vertically at intervals of 20 pixels. So, we can find uniform and nonuniform regions of the waveforms and define the period of uniform waveforms as an obstacle region. We can see that the algorithm is very useful for the robot to move avoiding obstacles.

• The Korean Journal of Nuclear Medicine
• /
• v.32 no.4
• /
• pp.365-373
• /
• 1998

Purpose: The conventional gamma camera is not ideal for scintimammography because of its large detector size (${\sim}500mm$ in width) causing high cost and low image quality. We are developing a small gamma camera dedicated for breast imaging. Materials and Methods: The small gamma camera system consists of a NaI (T1) crystal ($60 mm{\times}60 mm{\times}6 mm$) coupled with a Hamamatsu R3941 Position Sensitive Photomultiplier Tube (PSPMT), a resister chain circuit, preamplifiers, nuclear instrument modules, an analog to digital converter and a personal computer for control and display. The PSPMT was read out using a standard resistive charge division which multiplexes the 34 cross wire anode channels into 4 signals ($X^+,\;X^-,\;Y^+,\;Y^-$). Those signals were individually amplified by four preamplifiers and then, shaped and amplified by amplifiers. The signals were discriminated ana digitized via triggering signal and used to localize the position of an event by applying the Anger logic. Results: The intrinsic sensitivity of the system was approximately 8,000 counts/sec/${\mu}Ci$. High quality flood and hole mask images were obtained. Breast phantom containing $2{\sim}7 mm$ diameter spheres was successfully imaged with a parallel hole collimator The image displayed accurate size and activity distribution over the imaging field of view Conclusion: We have succesfully developed a small gamma camera using NaI(T1)-PSPMT and nuclear Instrument modules. The small gamma camera developed in this study might improve the diagnostic accuracy of scintimammography by optimally imaging the breast.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.15 no.3
• /
• pp.645-653
• /
• 2011

The aim of this study was to develop a Pulsed Magnetic Fields(PMFs) system which can produce effects locally and simulate muscular tissues equally. To evaluate the PMFs system we caused DOMS(Delayed Onset Muscle Soreness) to subjects in biceps of the arm. Then, we stimulated acupoint HT2 using PMFs(20 minutes) and TEAS(20 minutes) for 2 days. The other subjects did not stimulate. Then we checked body surface temperature in biceps of the arm. All subjects had an asymmetrical body surface temperature in biceps after exercise(Non-stimulation group=$2.00{\pm}1.16^{\circ}C$, TEAS group=$1.73{\pm}0.52^{\circ}C$, PMFs group=$1.48{\pm}0.51^{\circ}C$). After 1st stimulation all subjects had decreased temperature differences(Non-stimulation group=$1.37{\pm}0.71^{\circ}C$, TEAS group=$1.08{\pm}0.43^{\circ}C$, PMFs group=$1.23{\pm}0.15^{\circ}C$). PMFs group had a symmetry body surface temperature after 24 hours($0.05{\pm}0.06^{\circ}C$) and TEAS group had that after 48 hours($0.1{\pm}0.08^{\circ}C$). Non-stimulation group did not recovery after 48 hours($0.37{\pm}0.06^{\circ}C$). Therefore, PMFs on acupoint had an therapeutic effect in DOMS.