스마트 벤트 홀을 가진 조수석 에어백의 전개 시뮬레이션

김영관(Younggwan Kim),김권희(Kwonhee Kim),김형준(Hyungjun Kim),남영호(Youngho Nam) 한국자동차공학회 2013 한국자동차공학회 부문종합 학술대회 Vol.2013 No.5

The degree of protection from passenger airbag depends on the details of airbag deployment such as shape and size of the airbag, airbag pressure, vent hole size as well as the size of passenger. Larger passengers require higher airbag pressure than smaller passengers. Adaptive vent systems detect the weight of the passenger and controls the vent hole size with costly mechanisms. In this work, feasibility of an economic alternative system with pressure sensitive vent mechanism is explored. The pressure ? vent rate relation is explored via airbag deployment simulations with reference to drop tower tests with varying drop weight.

차량에서 도로 경사도의 실시간 측정 방법

김영관(Younggwan Kim),이지석(Jiseok Lee),위효성(Hyoseong Wi),박진일(Jinil Park),이종화(Jonghwa Lee) 한국자동차공학회 2012 한국자동차공학회 부문종합 학술대회 Vol.2012 No.5

Through the information about the tilt of road, Demanded power of the driving course on the vehicle can be predicted. Cause of prediction of the demanded power on vehicle, Driver can choose the best fuel-economy driving course, And it can expect to increase the fuel economy. For measuring the tilt of road, Using the clinometer is one of the safe method, But it needs so much time and efforts. For this reason, If measuring the tilt of road on vehicle is possible, Then it will be much easier. But measuring the tilt of road on vehicle does not have good accuracy, Because the signal which is measured on vehicle contains many kind of noise. In this paper, Propose the method of measurement about the tilt of road on vehicle and verify the proposed method through experiment.

도로 경사도의 실시간 측정 방법에 관한 연구

김영관(Younggwan Kim),박진일(Jinil Park),이종화(Jonghwa Lee) 한국생산제조학회 2013 한국생산제조학회지 Vol.22 No.6

차체의 스몰 오버랩 정면 충돌특성 향상을 위한 연구

김영관(Younggwan Kim),김권희(Kwonhee Kim),최선웅(Seonung Choi) (사)한국CDE학회 2016 한국 CAD/CAM 학회 학술발표회 논문집 Vol.2016 No.동계

기계학습 분산 환경을 위한 부하 분산 기법

김영관(Younggwan Kim),이주석(Jusuk Lee),김아정(Ajung Kim),홍지만(Jiman Hong) 한국스마트미디어학회 2021 스마트미디어저널 Vol.10 No.1

기계학습이 보편화되면서 기계학습을 활용한 응용 개발 또한 활발하게 이루어지고 있다. 또한 이러한 응용 개발을 지원하기 위한 기계학습 플랫폼 연구도 활발하게 진행되고 있다. 그러나 기계학습 플랫폼 연구가 활발하게 진행되고 있음에도 불구하고 기계학습 플랫폼에 적절한 부하 분산에 관한 연구는 아직 부족하다. 따라서 본 논문에서는 기계학습 분산 환경을 위한 부하 분산 기법을 제안한다. 제안하는 기법은 분산 서버를 레벨 해시 테이블 구조로 구성하고 각 서버의 성능을 고려하여 기계학습 작업을 서버에 할당한다. 이후 분산 서버를 구현하여 실험하고 기존 해싱 기법과 성능을 비교하였다. 제안하는 기법을 기존 해싱 기법과 비교하였을 때 평균 약 26%의 속도 향상을 보였고, 서버에 할당되지 못하고 대기하는 작업의 수가 약 38% 이상 감소함을 보였다. As the machine learning becomes more common, development of application using machine learning is actively increasing. In addition, research on machine learning platform to support development of application is also increasing. However, despite the increasing of research on machine learning platform, research on suitable load balancing for machine learning platform is insufficient. Therefore, in this paper, we propose a load balancing scheme that can be applied to machine learning distributed environment. The proposed scheme composes distributed servers in a level hash table structure and assigns machine learning task to the server in consideration of the performance of each server. We implemented distributed servers and experimented, and compared the performance with the existing hashing scheme. Compared with the existing hashing scheme, the proposed scheme showed an average 26% speed improvement, and more than 38% reduced the number of waiting tasks to assign to the server.

기본주파수와 성도길이의 상관관계를 이용한 HTS 음성합성기에서의 목소리 변환

유효근(Yoo, Hyogeun),김영관(Kim, Younggwan),서영주(Suh, Youngjoo),김회린(Kim, Hoirin) 한국음성학회 2017 말소리와 음성과학 Vol.9 No.1

The main advantage of the statistical parametric speech synthesis is its flexibility in changing voice characteristics. A personalized text-to-speech(TTS) system can be implemented by combining a speech synthesis system and a voice transformation system, and it is widely used in many application areas. It is known that the fundamental frequency and the spectral envelope of speech signal can be independently modified to convert the voice characteristics. Also it is important to maintain naturalness of the transformed speech. In this paper, a speech synthesis system based on Hidden Markov Model(HMM-based speech synthesis, HTS) using the STRAIGHT vocoder is constructed and voice transformation is conducted by modifying the fundamental frequency and spectral envelope. The fundamental frequency is transformed in a scaling method, and the spectral envelope is transformed through frequency warping method to control the speaker’s vocal tract length. In particular, this study proposes a voice transformation method using the correlation between fundamental frequency and vocal tract length. Subjective evaluations were conducted to assess preference and mean opinion scores(MOS) for naturalness of synthetic speech. Experimental results showed that the proposed voice transformation method achieved higher preference than baseline systems while maintaining the naturalness of the speech quality.

사물인식을 위한 딥러닝 모델 선정 플랫폼

이한솔(Hansol Lee),김영관(Younggwan Kim),홍지만(Jiman Hong) 한국스마트미디어학회 2019 스마트미디어저널 Vol.8 No.2

최근 컴퓨터 비전을 활용한 사물인식 기술이 센서 기반 사물인식 기술을 대체할 기술로 주목을 받고 있다. 센서 기반 사물인식 기술은 일반적으로 고가의 센서를 필요로 하기 때문에 기술이 상용화되기 어렵다는 문제가 있었다. 반면 컴퓨터 비전을 활용한 사물인식 기술은 고가의 센서 대신 비교적 저렴한 카메라를 사용할 수 있다. 동시에 CNN이 발전하면서 실시간 사물인식이 가능해진 이후 IoT, 자율주행자동차 등 타 분야에 활발하게 도입되고 있다. 그러나 사물 인식 모델을 상황에 알맞게 선택하고 학습시키기 위해서는 딥러닝에 대한 전문적인 지식을 요구하기 때문에 비전문가가 사물 인식 모델을 사용하기에는 어려움이 따른다. 따라서 본 논문에서는 딥러닝 기반 사물인식 모델들의 구조와 성능을 분석하고, 사용자가 원하는 조건의 최적의 딥러닝 기반 사물 인식 모델을 스스로 선정할 수 있는 플랫폼을 제안한다. 또한 통계에 기반한 사물 인식 모델 선정이 필요한 이유를 실험을 통해 증명한다. Recently, object recognition technology using computer vision has attracted attention as a technology to replace sensor-based object recognition technology. Sensor-based object recognition technology has a problem that it is difficult to commercialize the technology because an expensive sensor is required. On the other hand, since object recognition technology using computer vision can replace sensors with inexpensive cameras. Moreover, Real-time object recognition becomes possible because of the development of CNN, it is actively introduced into other fields such as IoT and autonomous vehicles. However, using the object recognition model requires expert knowledge on deep learning to select and learn the model, it is difficult for non-experts to use it. Therefore, in this paper, we analyze the structure of deep - learning - based object recognition models, and propose a platform that can automatically select a deep - running object recognition model based on a user s desired condition. We also show the reason why we need to select the object recognition model based on the statistics through experiments on the models.

L1-norm regularization을 통한 SGMM의 state vector 적응

구자현(Goo, Jahyun),김영관(Kim, Younggwan),김회린(Kim, Hoirin) 한국음성학회 2015 말소리와 음성과학 Vol.7 No.3

In this paper, we propose L1-norm regularization for state vector adaptation of subspace Gaussian mixture model (SGMM). When you design a speaker adaptation system with GMM-HMM acoustic model, MAP is the most typical technique to be considered. However, in MAP adaptation procedure, large number of parameters should be updated simultaneously. We can adopt sparse adaptation such as L1-norm regularization or sparse MAP to cope with that, but the performance of sparse adaptation is not good as MAP adaptation. However, SGMM does not suffer a lot from sparse adaptation as GMM-HMM because each Gaussian mean vector in SGMM is defined as a weighted sum of basis vectors, which is much robust to the fluctuation of parameters. Since there are only a few adaptation techniques appropriate for SGMM, our proposed method could be powerful especially when the number of adaptation data is limited. Experimental results show that error reduction rate of the proposed method is better than the result of MAP adaptation of SGMM, even with small adaptation data.

차량 실내 소음 주파수 특성 변경이 운전성 평가에 미치는 영향

안동환(Donghwan Ahn),이동현(Donghyun Lee),김영관(Younggwan Kim),위효성(Hyoseong Wi),이진우(Jinwoo Lee),박진일(Jinil Park),이종화(Jonghwa Lee),김범수(Beomsoo Kim),이석재(Seogjae Lee) 한국자동차공학회 2011 한국자동차공학회 학술대회 및 전시회 Vol.2011 No.11

Vehicle performance is improving day by day. So drivers want to higher quality of drivability, which include acceleration performance, interior noise, vibrations and etc. In particular, the vehicle interior noise which is a major factor in the evaluation drivability is important factor. In this study, According to the frequency of vehicle interior noise, vehicle interior noise were analyzed which based on sound pressure level. On the basis of this analysis, the interior noise was changed through sound equipment, such as a microphone and equalizer. This interior noise was evaluated by drivers and identified a correlation between subject’s responses and changing of the frequency. As a result of this research, frequency bands and sound pressure levels which affect evaluation drivability were presented to meet the needs of consumers about acceleration, noise and vibration.

한국어 text-to-speech(TTS) 시스템을 위한 엔드투엔드 합성 방식 연구

최연주(Choi, Yeunju),정영문(Jung, Youngmoon),김영관(Kim, Younggwan),서영주(Suh, Youngjoo),김회린(Kim, Hoirin) 한국음성학회 2018 말소리와 음성과학 Vol.10 No.1

A typical statistical parametric speech synthesis (text-to-speech, TTS) system consists of separate modules, such as a text analysis module, an acoustic modeling module, and a speech synthesis module. This causes two problems: 1) expert knowledge of each module is required, and 2) errors generated in each module accumulate passing through each module. An end-to-end TTS system could avoid such problems by synthesizing voice signals directly from an input string. In this study, we implemented an end-to-end Korean TTS system using Google’s Tacotron, which is an end-to-end TTS system based on a sequence-to-sequence model with attention mechanism. We used 4392 utterances spoken by a Korean female speaker, an amount that corresponds to 37% of the dataset Google used for training Tacotron. Our system obtained mean opinion score (MOS) 2.98 and degradation mean opinion score (DMOS) 3.25. We will discuss the factors which affected training of the system. Experiments demonstrate that the post-processing network needs to be designed considering output language and input characters and that according to the amount of training data, the maximum value of n for n-grams modeled by the encoder should be small enough.