• Journal of Information Processing Systems
• /
• v.14 no.5
• /
• pp.1176-1194
• /
• 2018

The object classification using the images' contents is a big challenge in computer vision. The superpixels' information can be used to detect and classify objects in an image based on locations. In this paper, we proposed a methodology to detect and classify the image's pixels' locations using enhanced bag of words (BOW). It calculates the initial positions of each segment of an image using superpixels and then ranks it according to the region score. Further, this information is used to extract local and global features using a hybrid approach of Scale Invariant Feature Transform (SIFT) and GIST, respectively. To enhance the classification accuracy, the feature fusion technique is applied to combine local and global features vectors through weight parameter. The support vector machine classifier is a supervised algorithm is used for classification in order to analyze the proposed methodology. The Pascal Visual Object Classes Challenge 2007 (VOC2007) dataset is used in the experiment to test the results. The proposed approach gave the results in high-quality class for independent objects' locations with a mean average best overlap (MABO) of 0.833 at 1,500 locations resulting in a better detection rate. The results are compared with previous approaches and it is proved that it gave the better classification results for the non-rigid classes.

• Journal of Oral Medicine and Pain
• /
• v.19 no.1
• /
• pp.125-136
• /
• 1994

Observation of elastic fiber's change of mouse TMJ due to several round factor, principally external stimulations, their influence on the TMJ structure's change and the analize of the consecutive evolution of the disease in most important. So, the author believe that the factor of TMJ feature is the elastic feature's change and it's the principal factor of the TMJ disease. For observation of the increase and disposition of elastic fiber that to regulate the elastic feature of tissue and allow it existence. For this propose, observation with histologic methods on 20mouse ICR of 3 days, 1 week, 2 weeks, 3 weeks and 4 weeks. The results were as follow : 1. In the early stage, the condyle of TMJ is originated from cartilage mass, and it's calcification is endochondral. 2. In the early stage, the disc is relatively thin and immature, but in the later stage the fiber is dense and the disposition is most functional. 3. Observation of the early stage, the elastic fiber is a thin fiber that to across antero- posterior direction, but in the later stage elastic fiber are developed, the disposition that in the early stage was perpendicular to articular surface, now in parallel. 4. The elastic fiber was observated most clearly in the retrodiscal tissue. 5. In conclusion, the elastic fiber is observed like a thin fiber 1 week from born, but the fiber to increase the weight and it dispose functionally, and 4 week from born, it can realize the normal function.

• KIPS Transactions on Software and Data Engineering
• /
• v.2 no.5
• /
• pp.341-346
• /
• 2013

In this paper, we propose a rule extraction method using a modified Fuzzy Min-Max (FMM) neural network. The suggested method supplements the hyperbox definition with a frequency factor of feature values in the learning data set. We have defined a relevance factor between features and pattern classes. The proposed model can solve the ambiguity problem without using the overlapping test process and the contraction process. The hyperbox membership function based on the fuzzy partitions is defined for each dimension of a pattern class. The weight values are trained by the feature range and the frequency of feature values. The excitatory features and the inhibitory features can be classified by the proposed method and they can be used for the rule generation process. From the experiments of sign language recognition, the proposed method is evaluated empirically.

• Journal of the Korean Society for Precision Engineering
• /
• v.18 no.1
• /
• pp.98-103
• /
• 2001

Machining is the commonly used process in the manufacturing of prototypes. This process offers several advantages, such as rigidity of the machine, precision of the machine, precision of the operation and specially a quick delivery. The weight and immobility of the machine support and immobilize the part during the operation. However, despite these advantages it shows, machining still presents several limitations. The immobilization, location and support of the part are referred to as fixturing or workholding and present the biggest challenge for time efficient machining. So it is important to select and design the appropriate fixturing assembly. This assembly depends on the complexity of the part and the tool paths and may require the construction of dedicated fixtures. With traditional techniques, the range of fixturable shapes is limited and the identification of suitable fixtures in a given setup involves complex reasoning. To solve this limitation and to apply the automation, this paper presents the Reference Free Part Encapsulation(RFPE) and implementation of the encapsulation system. The feature-based modeling system and the encapsulation system are implemented. The small part of which it is difficult to find out the appropriate fixturing assembly is made by this system.

• Journal of Institute of Control, Robotics and Systems
• /
• v.16 no.12
• /
• pp.1150-1158
• /
• 2010

In this study, the Polynomial-based Radial Basis Function Neural Networks is proposed as one of the recognition part of overall face recognition system that consists of two parts such as the preprocessing part and recognition part. The design methodology and procedure of the proposed pRBFNNs are presented to obtain the solution to high-dimensional pattern recognition problem. First, in preprocessing part, we use a CCD camera to obtain a picture frame in real-time. By using histogram equalization method, we can partially enhance the distorted image influenced by natural as well as artificial illumination. We use an AdaBoost algorithm proposed by Viola and Jones, which is exploited for the detection of facial image area between face and non-facial image area. As the feature extraction algorithm, PCA method is used. In this study, the PCA method, which is a feature extraction algorithm, is used to carry out the dimension reduction of facial image area formed by high-dimensional information. Secondly, we use pRBFNNs to identify the ID by recognizing unique pattern of each person. The proposed pRBFNNs architecture consists of three functional modules such as the condition part, the conclusion part, and the inference part as fuzzy rules formed in 'If-then' format. In the condition part of fuzzy rules, input space is partitioned with Fuzzy C-Means clustering. In the conclusion part of rules, the connection weight of pRBFNNs is represented as three kinds of polynomials such as constant, linear, and quadratic. Coefficients of connection weight identified with back-propagation using gradient descent method. The output of pRBFNNs model is obtained by fuzzy inference method in the inference part of fuzzy rules. The essential design parameters (including learning rate, momentum coefficient and fuzzification coefficient) of the networks are optimized by means of the Particle Swarm Optimization. The proposed pRBFNNs are applied to real-time face recognition system and then demonstrated from the viewpoint of output performance and recognition rate.

• The Journal of Korean Physical Therapy
• /
• v.14 no.1
• /
• pp.99-108
• /
• 2002

The purpose of this study was to research the clinical of cerebral palsy taking physical therapy at the department of physical therapy of various clinics at Kyungnam${\cdot}$Pusan. Among the subjects that was born from January, 1985 to June, 2000, 226 parents was answered to questionary of this study. The results of the study were as follows: 1) During the embryonic period, the cerebral palsied children above 37weeks were 114 subjects(50.9%) and there was 51 subjects(22.8%) between 28weeks to 31weeks and 32weeks to 36weeks. The children below 28weeks were 8 subjects(3.6%) and showed the lowest rate. As compared to the delievery methods, the normal delievery, cesserian section delievery, and forceps delievery was 124 subjects(55.1%), 81(36.0%), 16(7.1%), nad 4(1.8%). Among them the mormal delievery indicated the highest percentage. 2) Compared to the weight during birth time, the above 2500g of 121 subjects(55.3%) showed the highest rate and the 28 subjects(12.8%) had the birth weight of 1000to 1499g. There was 4 subjects(1.8%) below the 1000g. 3) Compared to the birth weight of the pregnancy period, the weight of the cerebral palsied children below 28weeks were 1000g to 1499g and showed the highest rate of 4 subjects(50%). The children between 28weeks to 31weeks and 32weeks to 36weeks were 1500 to 2499g, each 23(47.9%), and 28(54.9%) subject. The weight of the children of the above 37weeks were above 2500g and 94 subjects(87.4%). Therefore, if the period of pregnancy is short, the weight birth would indicate the lower weight than the weight of the other times(p<0.05). 4) The spastic type of the pregnant period had the highest rate and the period was the below 28weeks to 31weeks. The cerebral palsied children of athetoid and mixed type were 6 subjects(13.3%) and 5 subjects(31.9%) between 28 and 31weeks. The mixed type of them was each 15 (31.9%) and 33 (30.6%) subjects between 32 to 36 weeks and the above 37weeks. The mixed type showed a slightly high rate (p<0.05). 5) The spastic type indicated the highest rate in the weight of birth time and especially showed the high rate in the case of 1000 to 1499g. The mixed type indicated a slightly high rate of 17 subjects (25.8%) and 32 subjects (29.1%) in case of 1500 to 2499g and the above 2500g (p<0.05).

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.17 no.6
• /
• pp.105-111
• /
• 2017

In this paper, the analysis results are shown about several weights of Weighted K-Nearest Neighbor method, Recently, it is employed for the indoor positioning technologies using WiFi fingerprint which has been actively studied. In spite of the simplest feature, the W-KNN method shows comparable performance to another methods using WiFi fingerprint technology. So W-KNN method has employed in the existing indoor positioning system. It shows positioning error performance according to data preprocessing and weight factor, and the analysis on the weight is very important. In this paper, based on the real measured WiFi fingerprint data, the estimation error is analyzed and the performances are compared, for the case of data processing methods, of the weight of average, variance, and distance, and of the averaging several position of number K. These results could be practically useful to construct the real indoor positioning system.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.22 no.9
• /
• pp.1165-1171
• /
• 2018

CNN requires a large amount of computation and memory in the process of extracting the feature of the object. Also, It is trained from the network that the user has configured, and because the structure of the network is fixed, it can not be modified during training and it is also difficult to use it in a mobile device with low computing power. To solve these problems, we apply a pruning method to the pre-trained weight file to reduce computation and memory requirements. This method consists of three steps. First, all the weights of the pre-trained network file are retrieved for each layer. Second, take an absolute value for the weight of each layer and obtain the average. After setting the average to a threshold, remove the weight below the threshold. Finally, the network file applied the pruning method is re-trained. We experimented with LeNet-5 and AlexNet, achieved 31x on LeNet-5 and 12x on AlexNet.

• Journal of Institute of Control, Robotics and Systems
• /
• v.17 no.10
• /
• pp.1021-1028
• /
• 2011

The objective of this paper is to optimize the design parameters of a novel mechanism for a robotic knee orthosis. The feature of the proposed knee othosis is to drive a knee joint with independent actuation during swing and stance phases, which can allow an actuator with fast rotation to control swing motions and an actuator with high torque to control stance motions, respectively. The quadriceps device operates in five-bar links with 2-DOF motions during swing phase and is changed to six-bar links during stance phase by the contact motion to the patella device. The hamstring device operates in a slider-crank mechanism for entire gait cycle. The suggested kinematic model will allow a robotic knee orthosis to use compact and light actuators with full support during walking. However, the proposed orthosis must use additional linkages than a simple four-bar mechanism. To maximize the benefit of reducing the actuators power by using the developed kinematic design, it is necessary to minimize total weight of the device, while keeping necessary actuator performances of torques and angular velocities for support. In this paper, we use a SGA (Simple Genetic Algorithm) to minimize sum of total link lengths and motor power by reducing the weight of the novel knee orthosis. To find feasible parameters, kinematic constraints of the hamstring and quadriceps mechanisms have been applied to the algorithm. The proposed optimization scheme could reduce sum of total link lengths to half of the initial value. The proposed optimization scheme can be applied to reduce total weight of general multi-linkages while keeping necessary actuator specifications.

• The Journal of Korean Institute of Next Generation Computing
• /
• v.14 no.6
• /
• pp.15-29
• /
• 2018

This paper proposes an enhanced object tracking algorithm to compensate the lack of temporal information in existing particle swarm optimization based object trackers using the trajectory of the target object. The proposed scheme also enables the tracking and documentation of the location of an online updated set of distractions. Based on the trajectories information and the distraction set, a rule based approach with adaptive parameters is utilized for occlusion detection and determination of the target position. Compare to existing algorithms, the proposed approach provides more comprehensive use of available information and does not require manual adjustment of threshold values. Moreover, an effective weight adjustment function is proposed to alleviate the diversity loss and pre-mature convergence problem in particle swarm optimization. The proposed weight function ensures particles to search thoroughly in the frame before convergence to an optimum solution. In the existence of multiple objects with similar feature composition, this algorithm is tested to significantly reduce convergence to nearby distractions compared to the other existing swarm intelligence based object trackers.