• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.14 no.10
• /
• pp.4117-4135
• /
• 2020

Due to the angle of view, illumination and scene diversity, real-time dynamic face detection and recognition is no small difficulty in those unrestricted environments. In this study, we used the intrinsic correlation between detection and calibration, using a multi-task cascaded convolutional neural network(MTCNN) to improve the efficiency of face recognition, and the output of each core network is mapped in parallel to a compact Euclidean space, where distance represents the similarity of facial features, so that the target face can be identified as quickly as possible, without waiting for all network iteration calculations to complete the recognition results. And after the angle of the target face and the illumination change, the correlation between the recognition results can be well obtained. In the actual application scenario, we use a multi-camera real-time monitoring system to perform face matching and recognition using successive frames acquired from different angles. The effectiveness of the method was verified by several real-time monitoring experiments, and good results were obtained.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2022.10a
• /
• pp.97-99
• /
• 2022

For the early identification and treatment of colorectal cancer, accurate polyp segmentation is crucial. However, polyp segmentation is a challenging task, and the majority of current approaches struggle with two issues. First, the position, size, and shape of each individual polyp varies greatly (intra-class inconsistency). Second, there is a significant degree of similarity between polyps and their surroundings under certain circumstances, such as motion blur and light reflection (inter-class indistinction). U-net, which is composed of convolutional neural networks as encoder and decoder, is considered as a standard for tackling this task. We propose an updated U-net architecture replacing the encoder part with vision transformer network for polyp segmentation. The proposed architecture performed better than the standard U-net architecture for the task of polyp segmentation.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• v.12 no.4
• /
• pp.300-307
• /
• 2012

When the volume of data grows big, some simple tasks could become a significant concern. Nearest neighbor search is such a task which finds from a data set the k nearest data points to queries. Locality-sensitive hashing techniques have been developed for approximate but fast nearest neighbor search. This paper introduces the notion of locality-sensitive hashing and surveys the locality-sensitive hashing techniques. It categories them based on several criteria, presents their characteristics, and compares their performance.

• Journal of the Korean Data and Information Science Society
• /
• v.26 no.4
• /
• pp.857-864
• /
• 2015

By Wikipedia, data mining is the computational process of discovering patterns in huge data sets involving methods at the intersection of association rule, decision tree, clustering, artificial intelligence, machine learning. Clustering or cluster analysis is the task of grouping a set of objects in such a way that objects in the same group are more similar to each other than to those in other groups. The similarity measures being used in the clustering may be classified into various types depending on the characteristics of data. In this paper, we computed bounds for similarity measures based on the probabilistic interestingness measure with partially marginal probability such as Peirce I, Peirce II, Cole I, Cole II, Loevinger, Park I, and Park II measure. We confirmed the absolute value of Loevinger measure wasthe upper limit of the absolute value of any other existing measures. Ordering of other measures is determined by the size of concurrence proportion, non-simultaneous occurrence proportion, and mismatch proportion.

• Journal of the Korean Society for Nondestructive Testing
• /
• v.22 no.4
• /
• pp.422-429
• /
• 2002

The identification of ultrasonic flaw signals is a truly difficult task in the angle beam testing of welded joints due to non-relevant signals from the geometric reflectors such as weld roots and counter bores. This paper describes a new approach called "technique for identification of flaw signal using deconvolution(TIFD)" in order to identify the flaw signals in such a problematic situation. The concept of similarity function based on the deconvolution is introduced in the proposed approach. The "reference" signals from both flaws and geometric reflectors and test signals are acquired and normalized. The similarity functions are obtained by deconvolution of test signals with reference signals. The flaw signals could be identified by the patterns of similarity function. The initiative results show great potential of TIFD to distinguish notch comer signals from the geometric reflections.

• Korean Journal of Applied Biomechanics
• /
• v.18 no.2
• /
• pp.49-58
• /
• 2008

The purpose of this study was to identify effects of walking speed and a cognitive task during treadmill walking on gait variability. Experiments consisted of 5 different walking speeds(80%, 90%, 100%, 110% and 120% of preferred walking speed) with/without a cognitive task. 3D motion analysis system was used to measure subject's kinematic data. Temporal/spatial variables were selected for this study; stride time, stance time, swing time, step time, double support time, stride length, step length and step width. Two parameters were used to compare stride-to-stride variability with/without cognitive task. One is the coefficient of variance which is used to describe the amount of variability. The other is the detrended fluctuation analysis which is used to infer self-similarity from fluctuation of aspects. Results showed that cognitive task may influence stride-to-stride variability during treadmill walking. Further study is necessary to clarify this result.

• Speech Sciences
• /
• v.10 no.2
• /
• pp.287-301
• /
• 2003

The purpose of this study was to examine whether the same vowel sounds are perceived differently by the two local dialect speakers, Seoul dialect speakers (SDS) and Kyungnam dialect speakers (KDS), whose vowel systems differ each other. In the first experiment SDS and KDS heard vowels synthesized in vowel space with F1 by F2 and categorized them into one of 10 Korean monophthongs. The results showed that SDS and KDS perceived the synthesized vowels differently. For example, /$\varepsilon$ versus /e/ contrast, ${\o}$/, and /y/ are differentiated by SDS, whereas they are perceptually confused by KDS. We also observed that /i/ could not be perceived unless the vowel synthesis included F3 or higher formants. In the second experiment SDS and KDS performed the similarity rating task of 10 synthesized Korean monophthongs. Two-dimensional MDS solution based on the similarity rating scores was obtained for each dialect group. The first dimension can be named 'vowel advancement' and the second 'vowel height'. The comparison of the two MDS solutions showed that the overall psychological distances among the vowels are shorter in KDS than SDS and that especially the distance between /$\Lambda$/ and /i/ is shorter in KDS than SDS. The result suggested that perception or mental representation of vowels depends on the vowel system of the listener's dialect or language. Further research problems were discussed in the final section.

• Journal of KIISE:Software and Applications
• /
• v.29 no.4
• /
• pp.258-269
• /
• 2002

There have been many studies of automatic term recognition (ATR) and they have achieved good results. However, there are scopes to improve the performance of extracting terms still further by using the additional technical dictionaries. This paper focuses on the method for extracting terms using the hierarchy among technical dictionaries. Moreover, a statistical method based on frequencies, foreign words, and nested relations assists extracting terms which do not appear in dictionaries. Our method produces relatively good results for this task.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.10 no.12
• /
• pp.5464-5484
• /
• 2016

Few samples face recognition has become a highly challenging task due to the limitation of available labeled samples. As two popular paradigms in face image representation, sparse component analysis is highly robust while parts-based paradigm is particularly flexible. In this paper, we propose a probabilistic generative model to incorporate the strengths of the two paradigms for face representation. This model finds a common spatial partition for given images and simultaneously learns a sparse component analysis model for each part of the partition. The two procedures are built into a probabilistic generative model. Then we derive the score function (i.e. feature mapping) from the generative score space. A similarity measure is defined over the derived score function for few samples face recognition. This model is driven by data and specifically good at representing face images. The derived generative score function and similarity measure encode information hidden in the data distribution. To validate the effectiveness of the proposed method, we perform few samples face recognition on two face datasets. The results show its advantages.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.18 no.8
• /
• pp.2103-2118
• /
• 2024

Chinese herbal slices (CHS) automated recognition based on computer vision plays a critical role in the practical application of intelligent Chinese medicine. Due to the complexity and similarity of herbal images, identifying Chinese herbal slices is still a challenging task. Especially, easily-confused CHS have higher inter-class and intra-class complexity and similarity issues, the existing deep learning models are less adaptable to identify them efficiently. To comprehensively address these problems, a novel tiny easily-confused CHS dataset has been built firstly, which includes six pairs of twelve categories with about 2395 samples. Furthermore, we propose a ResNeSt-CHS model that combines multilevel perception fusion (MPF) and perceptive sparse fusion (PSF) blocks for efficiently recognizing easilyconfused CHS images. To verify the superiority of the ResNeSt-CHS and the effectiveness of our dataset, experiments have been employed, validating that the ResNeSt-CHS is optimal for easily-confused CHS recognition, with 2.1% improvement of the original ResNeSt model. Additionally, the results indicate that ResNeSt-CHS is applied on a relatively small-scale dataset yet high accuracy. This model has obtained state-of-the-art easily-confused CHS classification performance, with accuracy of 90.8%, far beyond other models (EfficientNet, Transformer, and ResNeSt, etc) in terms of evaluation criteria.