• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2020.11a
• /
• pp.347-350
• /
• 2020

본 논문에서는 영상 변환 기술인 이산웨이블릿변환(Discrete Wavelet Transform, DWT)를 딥러닝 기반의 네트워크로 구현한다. 딥러닝 기술 중에도 CNN 기반으로 네트워크를 설계하였으며, 본 DWT 네트워크는 해상도에 의존적이지 않은 계층들로만 구성된다. 데이터세트를 구성할 때 파이썬의 라이브러리를 사용하여 레이블 데이터세트를 구성한다. 128×128크기의 gray-scale 영상을 입력으로 사용하고 이에 대응하는 레이블 데이터세트를 구성하여 1-level DWT를 수행하는 네트워크의 학습을 진행한다. 역방향 변환도 네트워크 설계 후 데이터세트를 구성하여 학습을 진행한다. 학습이 완료된 1-level DWT 네트워크를 반복적으로 사용하여 Multi-level DWT 네트워크를 구성한다. 또한 양자화에 의한 간단한 영상압축 실험을 진행하여 DWT 네트워크의 성능과 압축 등의 응용분야에 활용할 수 있음을 보인다. 설계한 DWT 네트워크의 1-level 순방향 변환 성능은 42.18dB의 PSNR을 보였고, 1-level 역방향 변환 성능은 50.13dB의 PSNR을 보였다.

• The Transactions of the Korea Information Processing Society
• /
• v.1 no.1
• /
• pp.81-91
• /
• 1994

A DCT map and new cldebook design algorithm based on a two-dimension discrete cosine transform (2D-DCT) is presented for coder of image vector quantizer. We divide the image into smaller subblocks, then, using 2D DCT, separate it into blocks which are hard to code but it bears most of the visual information and easy to code but little visual information, and DCT map is made. According to this map, the significant features of training image are extracted by using the 2D DCT. A codebook is generated by partitioning the training set into a binary tree based on tree-structure. Each training vector at a nonterminal node of the binary tree is directed to one of the two descendants by comparing a single feature associated with that node to a threshold. Compared with the pairwise neighbor (PPN) and classified VQ(CVQ) algorithm, about 'Lenna' and 'Boat' image, the new algorithm results in a reduction in computation time and shows better picture quality with 0.45 dB and 0.33dB differences as to PNN, 0.05dB and 0.1dB differences as to CVQ respectively.

• Korean Journal of Human Ecology
• /
• v.15 no.3
• /
• pp.491-499
• /
• 2006

The purpose of this study is to measure the pizza purchasing behavioral characteristics of respondents and importances of factors affecting pizza purchase, to estimate the effects of attributes on choice of pizza restaurant, and to predict probability of selecting a particular pizza restaurant. The questionnaire consisted of two parts: The paired experimental profiles, purchasing behavior and importances of factors affecting pizza purchase. This study generated profiles of 16 hypothetical pizza restaurants based on seven attributes. The profiles comprised 16 discrete sets of variables, each of which had two levels. For this study, researcher randomly selected 150 university students as respondents. Twenty one students did not complete the survey instrument, resulting in a final sample size of 129. All estimations were carried out using frequencies, $X^2$, independent samples t-test, phreg procedure of SAS package. The results were as followed: Some purchasing behavioral characteristics and importances of factors affecting pizza purchase were significantly different by purchase frequency. Based on the estimated models developed for the two purchase frequency groups, the Chi-square statistics were significant at p<0.001. The parameter estimate for late delivery time with frequently purchase frequency group was highest, and the parameter estimate for price with frequently purchase frequency group was highest. The pizza restaurants that charged 20,000 won, offered 100% discount on eleventh pizza, promised to deliver pizza in 20 min, usually delivered the pizza as promised, offered 2 or more types of pizza crust, delivered steaming hot pizza, and did not offer a money-back guarantee which was favored by each of the two purchase frequency groups. The results from this study suggested that there was an opportunity to increase market share and profit by improving operations so that customers can receive discount and money-back guarantee simultaneously, and by reducing price, delivery time.

• Journal of Broadcast Engineering
• /
• v.24 no.1
• /
• pp.164-181
• /
• 2019

This paper introduces the analysis and development of digital holographic data codec technology to effectively compress hologram data. First, the generation method and data characteristics of the hologram standard data set provided by JPEG Pleno are introduced. We analyze energy compaction according to hologram generation method using discrete wavelet transform and discrete cosine transform. The quantization efficiency according to the hologram generation method is analyzed by applying uniform quantization and non-uniform quantization. We propose a transformation method quantization method suitable for hologram generation method through transform and quantization experiments. Finally, holograms are compressed using standard compression codecs such as JPEG, JPEG2000, AVC/H.264 and HEVC/H.265 and the results are analyzed.

• Journal of Internet of Things and Convergence
• /
• v.9 no.1
• /
• pp.9-17
• /
• 2023

Deep learning with large amount of computations is difficult to implement on micro-sized IoT devices or moblie devices. Recently, lightweight deep learning technologies have been introduced to make sure that deep learning can be implemented even on small devices by reducing the amount of computation of the model. Quantization is one of lightweight techniques that can be efficiently used to reduce the memory and size of the model by expressing parameter values with continuous distribution as discrete values of fixed bits. However, the accuracy of the model is reduced due to discrete value representation in quantization. In this paper, we introduce various quantization techniques to correct the accuracy. We selected APoT and EWGS from existing quantization techniques, and comparatively analyzed the results through experimentations The selected techniques were trained and tested with CIFAR-10 or CIFAR-100 datasets in the ResNet model. We found out problems with them through experimental results analysis and presented directions for future research.

• Journal of the Korean Chemical Society
• /
• v.42 no.6
• /
• pp.646-651
• /
• 1998

The molecular structure of sulfur compounds and the retention relationship are studied by gas chromatography. Analyzed sulfur compounds are, hydrogen sulfide, sulfur dioxide, carbon disulfide, ethyl mercaptan, dimethyl sulfide, iso-propyl mercaptan, normal propyl mercaptan, ethyl methyl sulfide, tert-butyl mercaptan, tetrahydrothiophene, thiophene, and 2-chlorothiophene. Multiple linear regression explains the retention relationship of molecular descriptors. In GC the temperature program is 30$^{\circ}C$ held for 10.5 min, and then increased to 150$^{\circ}C$ at a rate 15$^{\circ}C$/min. Predicted equation for relative retention time (RRT) using SAS program is as follows; $RRT=0.121bp+14.39dp-8.94dp^2+0.0741sqmw-35.78\; (N=8,\; R^2=0.989, \;Variance=0.175,\;F=66.21)$. RRTs are function of boiling point, the square root of molecular weight, molecular dipole moment, and boiling point effects mostly on RRT. The RRT is maximized at the molecular dipole moment of 0.805D, when using nonpolar columns. The planar and highly symmetric compounds are eluted slowly. The square, of correlation coefficient $(R^2)$ using SAS program, is 0.989, and the variance is 0.175 in training sets. For three sulfur compounds, the variance between observed RRTs and predicted RRTs is 0.432 in testing sets.

• Journal of Digital Contents Society
• /
• v.19 no.4
• /
• pp.821-827
• /
• 2018

Affective computing has gained increasing interest in the recent years with the development of potential applications in Human computer interaction (HCI) and healthcare. Although momentous research has been done on human emotion recognition, however, in comparison to speech and facial expression less attention has been paid to physiological signals. In this paper, Electroencephalogram (EEG) signals from different brain regions were investigated using modified wavelet energy features. For minimization of redundancy and maximization of relevancy among features, mRMR algorithm was deployed significantly. EEG recordings of a publically available "DEAP" database have been used to classify four classes of emotions with Multi class Support Vector Machine. The proposed approach shows significant performance compared to existing algorithms.

• KIPS Transactions on Software and Data Engineering
• /
• v.10 no.1
• /
• pp.9-18
• /
• 2021

Although Generative Adversarial Networks (GANs) have gained great popularity in computer vision and related fields, generating audio signals independently has yet to be presented. Unlike images, an audio signal is a sampled signal consisting of discrete samples, so it is not easy to learn the signals using CNN architectures, which is widely used in image generation tasks. In order to overcome this difficulty, GAN researchers proposed a strategy of applying time-frequency representations of audio to existing image-generating GANs. Following this strategy, we propose an improved method for increasing the fidelity of synthesized audio signals generated by using GANs. Our method is demonstrated on a public speech dataset, and evaluated by Fréchet Inception Distance (FID). When employing our method, the FID showed 10.504, but 11.973 as for the existing state of the art method (lower FID indicates better fidelity).