• 한국멀티미디어학회논문지
• /
• 제22권7호
• /
• pp.749-758
• /
• 2019

This paper proposes a data augmentation algorithm to improve the performance of DNN(Deep Neural Network) based speech enhancement. Many deep learning models are exploring algorithms to maximize the performance in limited amount of data. The most commonly used algorithm is the data augmentation which is the technique artificially increases the amount of data. For the effective data augmentation algorithm, we used a formant enhancement method that assign the different weights to the formant frequencies. The DNN model which is trained using the proposed data augmentation algorithm was evaluated in various noise environments. The speech enhancement performance of the DNN model with the proposed data augmentation algorithm was compared with the algorithms which are the DNN model with the conventional data augmentation and without the data augmentation. As a result, the proposed data augmentation algorithm showed the higher speech enhancement performance than the other algorithms.

• Maxillofacial Plastic and Reconstructive Surgery
• /
• 제38권
• /
• pp.37.1-37.5
• /
• 2016

Background: The dimensions and shape of the forehead determine the esthetics of the upper third of the face. Korean young people consider a broad and smooth, rounded forehead more attractive. As a result, frontal augmentation becomes more popular in patients with dentofacial deformities. Various surgical procedures and materials have been used in frontal augmentation surgery, with associated advantages and disadvantages. Silicone is a good candidate for frontal augmentation. The author presents two cases of esthetic frontal augmentation with a prefabricated silicone implant in female patients with dentofacial deformities. Case presentation: In case 1, a 24-year-old female patient underwent frontal augmentation surgery with simultaneous maxillomandibular and zygomatic osteotomies to correct facial asymmetry. A silicone implant was fabricated preoperatively using a positive template stone mold of her forehead. In case 2, a 23-year-old female patient underwent total facial contouring surgery including frontal augmentation for improved facial esthetics. A computed tomography (CT)-guided rapid prototype (RP) model was used to make the silicone implants. The operative procedure was safe and simple, and the silicone implants were reliable for a larger degree of frontal augmentation. Six months later, both patients had recovered from the surgery and were satisfied with their frontal shape and projection. Conclusions: Frontal augmentation with silicone implants can be an effective adjuvant strategy to improve facial esthetics in patients with a flat and narrow forehead who undergo orthognathic reconstruction or total facial contouring surgery.

• Journal of Periodontal and Implant Science
• /
• 제38권2호
• /
• pp.207-214
• /
• 2008

Purpose: The aim of augmentation of the alveolar ridge is to restore absorbed alveolar ridges for future implant site or esthetic prosthodontic restoration. The present clinical report describes the anterior maxillary augmentation cases using a soft tissue rotated palatal flap, and considers various problems of before and after surgery. Method: First & second patients were treated by vascularized interpositional periosteal-connective tissue(VIP-CT) flap for horizontal soft tissue augmentation. Especially second patient was progressed with bone grafting at the same time. Third patient was treated by the same flap with bone graft and implant placement in single tooth missing premaxillary area. Result: The obtained horizontal augmentation width measured $0.5{\sim}2.7\;mm$. Conclusion: This technique constitutes a viable approach for augmentation the anterior sector of alveolar ridge with the placement of dental implants. But it needs correct diagnosis preparation and careful surgery skill.

• 전기전자학회논문지
• /
• 제26권2호
• /
• pp.176-185
• /
• 2022

데이터 증강이란 다양한 데이터 변환 및 왜곡을 통해 데이터셋의 크기와 품질을 개선하는 기법으로, 기계학습 모델의 과적합 문제를 해결하기 위한 대표적인 접근법이다. 그러나 심층학습 이미지 생성 모델인 GAN 기반 모델에서 데이터 증강을 적용하면 생성된 이미지에 데이터 변환과 왜곡이 반영되는 증강 누출 문제가 발생하여 생성 이미지의 품질이 하락한다. 이러한 문제를 해결하기 위해 본 논문에서는 데이터 증강의 종류와 수에 관계없이 증강 누출을 방지하는 기법을 제안한다. 증강 누출의 발생 조건을 분석하였으며, 보조적인 데이터 증강 작업 분류기를 GAN 모델에 적용하여 증강 누출을 방지하였다. 정성적 정량적 평가를 통해 제안된 기법을 적용하면 증강 누출이 발생하지 않음을 보이고 추가적으로 생성 이미지의 품질을 향상시키며 기존 기법과 비교하여 발전된 성능을 보임을 입증하였다.

• 한국가시화정보학회:학술대회논문집
• /
• 한국가시화정보학회 2004년도 추계학술대회 논문집
• /
• pp.22-25
• /
• 2004

The present paper investigated the correlation between acoustic pressure and heat transfer augmentation in acoustic fields. The acoustic pressure predicted by numerical work and compared with the augmentation ratio of heat transfer coefficient was experimentally measured. Also, particle image velocimetry(PIV) was used for the visualization of velocity vectors and kinetic energy distribution inside liquid region. For the numerical work, SVS programed with Fortran language and based on a coupled FE-BEM was used. Results of the present study, the acoustic pressure is increased by $60\%$ and the largest augmentation of heat transfer about $28\%$ was measured. Finally, the profiles of acoustic pressure is consistent with that of augmentation of heat transfer. It is concluded that a correlation exists between the acoustic pressure and the heat transfer augmentation.

• ETRI Journal
• /
• 제44권2호
• /
• pp.327-338
• /
• 2022

Classification models for human action recognition require robust features and large training sets for good generalization. However, data augmentation methods are employed for imbalanced training sets to achieve higher accuracy. These samples generated using data augmentation only reflect existing samples within the training set, their feature representations are less diverse and hence, contribute to less precise classification. This paper presents new data augmentation and action representation approaches to grow training sets. The proposed approach is based on two fundamental concepts: virtual video generation for augmentation and representation of the action videos through robust features. Virtual videos are generated from the motion history templates of action videos, which are convolved using a convolutional neural network, to generate deep features. Furthermore, by observing an objective function of the genetic algorithm, the spatiotemporal features of different samples are combined, to generate the representations of the virtual videos and then classified through an extreme learning machine classifier on MuHAVi-Uncut, iXMAS, and IAVID-1 datasets.

• Communications for Statistical Applications and Methods
• /
• 제30권3호
• /
• pp.343-354
• /
• 2023

Recently, there have been many improvements in general language models using architectures such as GPT-3 proposed by Brown et al. (2020). Nevertheless, training complex models can hardly be done if the number of data is very small. Data augmentation that addressed this problem was more than normal success in image data. Image augmentation technology significantly improves model performance without any additional data or architectural changes (Perez and Wang, 2017). However, applying this technique to textual data has many challenges because the noise to be added is veiled. Thus, we have developed a novel method for performing data augmentation on text data. We divide the data into signals with positive or negative meaning and noise without them, and then perform data augmentation using k-doc augmentation to randomly combine signals and noises from all data to generate new data.

• Wind and Structures
• /
• 제19권2호
• /
• pp.199-217
• /
• 2014

Reported experimental and computational fluid dynamic (CFD) studies have demonstrated significant power augmentation of diffuser shrouded horizontal axis micro wind turbine compared to bare turbine. These studies also found the degree of augmentation is strongly dependent on the shape and geometry of the diffuser such as length and expansion angle. However study flow field over the rotor blades in shrouded turbine has not received much attention. In this paper, CFD simulations of an experimental diffuser shrouded micro wind turbine have been carried out with the aim to understand the mechanisms underpinning the power augmentation phenomenon. The simulations provide insight of the flow field over the blades of bare wind turbine and of shrouded one elucidating the augmentation mechanisms. From the analysis, sub-atmospheric back pressure leading to velocity augmentation at the inlet of diffuser and lowering the static pressure on blade suction sides have been identified as th dominant mechanisms driving the power augmentation. And effective augmentation was achieved for ${\lambda}$ above certain value. For the case turbine it is ${\lambda}$ greater than ${\approx}2$.

• Archives of Plastic Surgery
• /
• 제33권2호
• /
• pp.155-160
• /
• 2006

Although several reports have been introduced about dual plane augmentation mammaplasty, the description of periareolar approach dual plane augmentation mammaplasty was few. This article describes specific characteristics, and different classification and techniques for the periareolar dual plane breast augmentation while postoperative scars resulted from inframammary crease approach caused complaints. A total of 124 patients(248 breasts) had periareolar dual plane augmentation surgery from 1998 to 2004. Anatomic implants were used in 43 cases. Most of the patients were satisfied with the outcomes of periareolar dual plane augmentation. Periareolar dual plane augmentation mammaplasty adjusts implant and tissue relationships to ensure adequate soft-tissue coverage while optimizing implant-breast parenchymal dynamics to offer increased benefits and fewer faults compared to a single pocket location in a wide range of breast types with minimal scars. Two types of dual plane classifications are discussed in this study for the periareolar approach exclusively. The boundaries of retroglandular dissection remain constant, as the costal origin of pectoralis major are divided. Type A dual plane implies that the inferior edge of pectoralis muscle lies below the inferior areolar border, and type B dual plane implies that the inferior edge lies above the superior areolar border.

• Archives of Plastic Surgery
• /
• 제34권1호
• /
• pp.99-104
• /
• 2007

Purpose: Subfascial augmentation mammaplasty was introduced by Dr. Graf in 2000. Subfascial placement of breast implants for augmentation was advocated as an option that has some of the advantages of both the subpectoral and subglandular placement while minimizing the disadvantages of each. The clinical experiences of 23 breast augmentations in the subfascial placement are reported. The indications for this technique are proposed. The incidence of complications is described from clinical experiences and compared with that of other methods. Methods: From January of 2004 through December of 2005, 23 patients underwent periareolar subfascial augmentation mammaplasty. The mean postoperative follow-up time was 8 months. Results: In comparing the results of the subpectoral augmentation group(57 patients) with those of the dual plane(124 patients) and subfascial groups(23 patients), the total rate of complications didn't represented the significant difference. The benefits of this technique include avoiding hematoma(as seen in the dual plane) and muscle action(in the subpectoral), and minimizing postoperative chest pain(inherent to subpectoral), and the ability to correct ptosis. And also this subfascial technique can be used for changing the plane from submuscular to subfascial in case of the reoperations. Conclusion: We're thinking that the periareolar subfascial augmentation mammaplasty would be the very useful tool for the primary and secondary breast augmentations.