• The Journal of the Acoustical Society of Korea
• /
• v.35 no.1
• /
• pp.74-82
• /
• 2016

In this paper, a new automatic speech recognition (ASR) was proposed where ultrasonic doppler signals were used, instead of conventional speech signals. The proposed method has the advantages over the conventional speech/non-speech-based ASR including robustness against acoustic noises and user comfortability associated with usage of the non-contact sensor. In the method proposed herein, 40 kHz ultrasonic signal was radiated toward to the mouth and the reflected ultrasonic signals were then received. Frequency shift caused by the doppler effects was used to implement ASR. The proposed method employed multi-channel ultrasonic signals acquired from the various locations, which is different from the previous method where single channel ultrasonic signal was employed. The PCA(Principal Component Analysis) coefficients were used as the features of ASR in which hidden markov model (HMM) with left-right model was adopted. To verify the feasibility of the proposed ASR, the speech recognition experiment was carried out the 60 Korean isolated words obtained from the six speakers. Moreover, the experiment results showed that the overall word recognition rates were comparable with the conventional speech-based ASR methods and the performance of the proposed method was superior to the conventional signal channel ASR method. Especially, the average recognition rate of 90 % was maintained under the noise environments.

• Journal of KIISE:Computer Systems and Theory
• /
• v.26 no.9
• /
• pp.1166-1176
• /
• 1999

연속 매체, 특히 비디오 데이타에 대한 일반 사용자 연산에는 재생뿐만 아니라 임의 속도 탐색 연산, 정지 연산, 그리고 그 외 다양한 연산이 있다. 이 연산 중에서 원하는 화면을 빨리 찾는 데에 유용한 고속 전진(FF: fast-forward)과 고속 후진(FB: fast-backward)은 재생 연산과는 달리 비순차적인 디스크 접근을 요구한다. 이러한 경우에 디스크 부하가 균등하지 않으면 일부 디스크에 접근이 편중되어 서비스 품질이 떨어진다. 본 논문에서는 디스크 배열을 이용한 저장 시스템에서 디스크 접근을 고르게 분산시키기 위하여 '소수 라운드 로빈(PRR: Prime Round Robin)' 방식으로 연속 매체를 디스크에 배치하는 기법에서 문제가 됐던 낭비된 디스크 저장 공간을 신뢰도 향상을 위해서 사용하는 '그룹화된 패리티를 갖는 소수 라운드 로빈(PRRgp: PRR with Grouped Parities)' 방식을 제안한다. 이 기법은 PRR 기법처럼 임의 속도 검색 연산에 있어서 디스크 배열을 구성하는 모든 디스크의 부하를 균등하게 할뿐만 아니라 낭비됐던 디스크 저장 공간에 신뢰도를 높이기 위한 패리티 정보를 저장함으로서 신뢰도를 향상시킬 수 있다. 신뢰도 모델링 방법으로 조합 모델과 마르코프 모델을 이용해서 결함발생율과 결함복구율을 고려한 신뢰도를 산출하고 비교.분석한다. PRR 기법으로 연속 매체를 저장하고 낭비되는 공간에 패리티 정보를 저장할 경우에 동시에 두 개 이상의 결함 발생 시에 그 결함으로부터 복구가 불가능하지만 PRRgp 기법에서는 약 30% 이상의경우에 대해서 동시에 두 개의 결함 발생 시에 저장한 패리티 정보를 이용한 복구가 가능할 뿐만 아니라 패리티 그룹의 수가 두 개 이상인 경우에는 두 개 이상의 결함에 대해서도 복구가 가능하다.Abstract End-user operations on continuous media (say video data) consist of arbitrary-rate search, pause, and others as well as normal-rate play. FF(fast-forward) / FB(fast-backward) among those operations are desirable to find out the scene of interest but they require non-sequential access of disks. When accesses are clustered to several disks without considering load balance, high quality services in playback may not be available. In this paper, we propose a new disk placement scheme, called PRRgp(Prime Round Robin with Grouped Parities), with enhanced reliability by using the wasted disk storage space in an old one(PRR: Prime Round Robin), in which continuous media are placed on a disk array based storage systems to distribute disk accesses uniformly. The PRRgp can not only achieve load balance of disks consisting of a disk array under arbitrary-rate search like PRR, but also improve reliability by storing parity information on the wasted disk space appropriately. We use combinatorial and Markov models to evaluate the reliability for a disk array and to analyze the results. When continuous media like PRR are placed and parity information on the wasted disk space is stored, we cannot tolerate more than two simultaneous faults. But they can be recovered by using stored parity information for about 30 percent as a whole in case of PRRgp presented in this paper. In addition, more than two faults can be tolerated in case there are more than two parity groups.

• Journal of the Korean Geographical Society
• /
• v.47 no.1
• /
• pp.98-120
• /
• 2012

The main objective of this study is to conduct subnational population projections of Korea based on a multiregional cohort-component method. This objective is accomplished by: (i) establishing a viable framework to implement the multiregional cohort-component method with reference to the Markov chain model and Rogers' multiregional population projection model; (ii) applying the established framework to subnational population projections of Korea, 2005~2030. The main results are twofold. First, the proposed method turns out to be highly valid in a methodological sense, which is seen from a high level of coincidence between the estimated and the observed. Second, the projection results turn out to be highly useful in the sense that interregional migration flow matrices are resulted for projection periods. The projected migration flows are expected to provide invaluable information for an understanding of future population change and for a formulation of policy alternatives. This study is strongly inspired by the multiregional perspective emphasizing the evolution of multiple regional populations interconnected by interregional migration flows rather than the overall national change.

• Journal of KIISE:Information Networking
• /
• v.30 no.5
• /
• pp.581-594
• /
• 2003

Recently, a location management is being more important in mobile communication systems due to an explosive increase of mobile users. Current systems have used a concept of location area. Based on this concept, a mobile user performs a location update whenever it moves into a new location area. However, this scheme can not avoid unnecessary location updates when a mobile user moves around with high movement rate as compared to call arrival rate. That results in tremendous location management cost. To overcome this drawback, our proposal divides service areas into two sets: One is a set of areas that mobile users move with high speed and another is a set of areas that they move with low speed. After establishing these two sets, this paper employs different location tracking schemes for each sets. Generally, most mobile users with high speed have a low CMR and a regular direction until they arrive at their destination. Using such the moving behavior, systems can predict a mobile user's next location area in advance. When the mobile user moves into the predicted location, our proposal does not perform a location update. Thus, it can reduce overall location management cost. The Markov model is used to analyze the performance of our proposal. Using the model, this paper compares our proposal with IS-41 and TLA. The analytic results show that as CMR grows lower, an overall cost of our proposal becomes less, particularly if a mobile user frequently moves into the specific location are predicted by mobile systems. Also, our proposal has a better performance than other two schemes when the communication cost between HLR and VLR is high.

• Journal of Korea Water Resources Association
• /
• v.54 no.12
• /
• pp.1317-1328
• /
• 2021

This study aims to provide a predictive model based on climate models for simulating continuous daily rainfall sequences by combining bias-correction and spatio-temporal downscaling approaches. For these purposes, this study proposes a combined modeling system by applying conditional Copula and Multisite Non-stationary Hidden Markov Model (MNHMM). The GloSea5 system releases the monthly rainfall prediction on the same day every week, however, there are noticeable differences in the updated prediction. It was confirmed that the monthly rainfall forecasts are effectively updated with the use of the Copula-based bias-correction approach. More specifically, the proposed bias-correction approach was validated for the period from 1991 to 2010 under the LOOCV scheme. Several rainfall statistics, such as rainfall amounts, consecutive rainfall frequency, consecutive zero rainfall frequency, and wet days, are well reproduced, which is expected to be highly effective as input data of the hydrological model. The difference in spatial coherence between the observed and simulated rainfall sequences over the entire weather stations was estimated in the range of -0.02~0.10, and the interdependence between rainfall stations in the watershed was effectively reproduced. Therefore, it is expected that the hydrological response of the watershed will be more realistically simulated when used as input data for the hydrological model.

• The Journal of the Acoustical Society of Korea
• /
• v.40 no.5
• /
• pp.530-536
• /
• 2021

Recently, due to the development of deep learning, end-to-end speech recognition, which directly maps graphemes to speech signals, shows good performance. Especially, among the end-to-end models, conformer shows the best performance. However end-to-end models only focuses on the probability of which grapheme will appear at the time. The decoding process uses a greedy search or beam search. This decoding method is easily affected by the final probability output by the model. In addition, the end-to-end models cannot use external pronunciation and language information due to structual problem. Therefore, in this paper conformer with lexicon transducer is proposed. We compare phoneme-based model with lexicon transducer and grapheme-based model with beam search. Test set is consist of words that do not appear in training data. The grapheme-based conformer with beam search shows 3.8 % of CER. The phoneme-based conformer with lexicon transducer shows 3.4 % of CER.

• Journal of Korean Society of Forest Science
• /
• v.89 no.5
• /
• pp.655-665
• /
• 2000

Taking account of the structural variation on species composition by topography, the successional trends were comparatively analyzed for the three topographic positions (valley, mid-slope, and ridge) in the natural deciduous forest of Mt. Chumbong area. The analysis was based upon the subsequent process of generation replacement by understory saplings and seedlings over the overstory trees which will be eventually fallen down. This study adopted the plot sampling method, establishing twenty $20m{\times}20m$ quadrats and collecting vegetation and site data on each different topographic position. The transition matrix model, which was modified from the mathematical theory of Markov chain, was employed to analyze the successional trends and thereafter to predict the overstory species composition in the future for each different topographic position. In valley, the simulation indicated the remarkable decrease in the proportion of species composition of present dominants Quercus mongolica and Fraxinus mandshurica from current 23% and 21% to around 4% of each at the steady state, which is predicted to take less than 200 years. On the other hand, the proportion of such species as Abies holophylla, Acer mono, Tilia amurensis, and Ulmus laciniata will increase at the steady state. In mid-slope, the result showed the remarkable decrease in the proportion of Juglans mandshurica, Kalopanax pictus, and Tilia amurensis from current 15%, 8%, and 15% to 2%, 1%, and 5%, respectively, at steady state predicted to take more than 250 years. In ridge, the current dominant Quercus mongolica was predicted to be decreased dramatically from 58% to 8% at steady state which could be achieved about 200 years. On the contrary, the proportion of Acer mono and Tilia amurensis will be increased from current 4% and 3% to more than 20% and 40%, respectively, at the steady state. Overall results suggested that the study forest is more likely seral rather than climax community. Even though a lot of variation is inevitable due to various kinds of site and vegetation development, the study forest is considered to be more than 200 years away from the steady state or climax in terms of overstory species composition.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.49 no.3
• /
• pp.51-59
• /
• 2012

This paper suggests an efficient method which tracks articulated human body modeled with markov network using disparity map derived from stereo images. The conventional methods which only use color information to calculate likelihood for energy function tend to fail when background has same colors with objects or appearances of object are changed during the movement. In this paper, we present a method evaluating likelihood with both disparity information and color information to find human body parts. Since the human body part are cylinder projected to rectangles in 2D image plane, we use the properties of distribution of disparity of those rectangles that do not have discontinuous distribution. In addition to that we suggest a conditional-messages-update that is able to reduce unnecessary message update of belief propagation. Since the message update has comprised over 80% of the whole computation in belief propagation, the conditional-message-update yields 9~45% of improvements of computational time. Furthermore, we also propose an another speed up method called three dimensional dynamic models assumed the body motion is continuous. The experiment results show that the proposed method reduces the computational time as well as it increases tracking accuracy.

• Proceedings of the Korean Information Science Society Conference
• /
• 2005.07b
• /
• pp.433-435
• /
• 2005

임베디드 시스템은 다수의 디바이스를 컨트롤하여 시스템의 목적을 수행한다. 최근 임베디드 시스템의 요구사항이 증가함에 따라 하나의 임베디드 소프트웨어가 컨트롤하는 디바이스의 종류가 다양해지고 수도 증가하는 추세이다. 다수의 디바이스를 가지고 있는 임베디드 시스템에서 시스템의 신뢰도는 각 디바이스의 신뢰도에 많은 영향을 받는다. 본 논문은 임베디드 시스템의 신뢰도를 측정하기 위해서 통계적 신뢰도 측정 방법 중 한 가지인 마르코프 체인을 이용한 방법을 제안한다. 마르코프 체인은 여러 분야에서 복잡한 시스템을 단순화하여 모델링하고 과거의 변화를 토대로 미래를 예측할 수 있는 방법을 제공한다. 또한 전체 시스템의 확률을 행렬로 계측할 수 있는 방법을 가지고 있어 특정 부분의 확률이 전체 시스템의 확률에 미치는 영향을 산술적으로 계산할 수 있는 장점을 가지고 있다. 본 논문에서 제안하는 임베디드 소프트웨어 마르코프 체인은 테스트 대상 소스를 분석하여 디바이스를 컨트롤하는 루틴과 에러를 핸들링하는 루틴, 일반적인 루틴으로 나누어 각각을 상태로 정의한다. 정의한 각 상태간의 전이는 통계적으로 측정한 디바이스 신뢰도를 확률로 표현한다. 마르코프 체인을 이용하여 임베디드 시스템의 신뢰도를 측정하기 위한 시스템은 소스 분석기와 신뢰도 측정기로 나누어 설계한다. 소스 분석기는 테스트 대상이 되는 소스와 디바이스 드라이버 라이블러리 테이블을 입력으로 하고 소프트웨어의 마르코프 체인을 출력으로 한다 마르코프 체인은 행렬로 표현하고 연산하여 시스템의 신뢰도를 측정한다. 제안하는 시스템의 신뢰도 측정 방법은 부분이 가지고 있는 신뢰도가 전체 신뢰도에 미치는 영향을 산술적으로 측정할 수 있어 시스템이 요구하는 신뢰도에 접근할 수 있는 방법과 근거를 제공하는 장점이 있다.소시키는 장점을 갖는다.것으로 조사되었으며 40대 이상의 연령층은 점심비용으로 더 많은 지출을 하고 있는 것으로 나타났다. 4) 끼니별 한식에 대한 선호도는 아침식사의 경우가 가장 높았으며, 이는 40대와 50대에서 높게 나타났다. 점심 식사로 가장 선호되는 음식은 중식, 일식이었으며 저녁 식사에서 가장 선호되는 메뉴는 전 연령층에서 일식, 분식류 이었으며, 한식에 대한 선택 정도는 전 연령층에서 매우 낮게 나타났다. 5) 각 연령층에서 선호하는 한식에 대한 조사에서는 된장찌개가 전 연령층에서 가장 높은 선호도를 나타내었고, 김치는 40대 이상의 선호도가 30대보다 높게 나타났으며, 흥미롭게도 30세 이하의 선호도는 30대보다 높게 나타났다. 그 외에도 떡과 죽에 대한 선호도는 전 연령층에서 낮게 조사되었다. 장아찌류의 선호도는 전 연령대에서 낮았으며 특히 30세 이하에서 매우 낮게 조사되었다. 한식의 맛에 대한 만족도 조사에서는 연령이 올라갈수록 한식의 맛에 대한 만족도는 낮아지고 있었으나, 한식의 맛에 대한 만족도가 높을수록 양과 가격에 대한 만족도는 높은 경향을 나타내었다. 전반적으로 한식에 대한 선호도는 식사 때와 식사 목적에 따라 연령대 별로 다르게 나타나고 있으나, 선호도는 성별이나 세대에 관계없이 폭 넓은 선호도를 반영하고 있으며, 이는 대학생들을 대상으로 하는 연구 등에서도 나타난바 같다. 주 5일 근무제의 확산과 초 중 고생들의 토요일 휴무와 더불어 여행과 엔터테인먼트산업은 더욱 더 발전을 거듭하고 있으며, 외식은 여행과 여가 활동의 필수적인 요소로써 그 역할을 일조하고 있다. 이와 같은 여가시간의 증가는 독신자들에게는 좀더 많은 여유시간을 가족을 이루고 있는 가족구성원들에게는 가족과의 유대를 강화하는 휴식과 오락의 소비 트렌드를 창출시켰

• The Journal of Korea Institute of Information, Electronics, and Communication Technology
• /
• v.14 no.2
• /
• pp.134-139
• /
• 2021

Currently, the autonomous vehicle market is commercializing a level 3 autonomous vehicle, but it still requires the attention of the driver. After the level 3 autonomous driving, the most notable aspect of level 4 autonomous vehicles is vehicle stability. This is because, unlike Level 3, autonomous vehicles after level 4 must perform autonomous driving, including the driver's carelessness. Therefore, in this paper, we propose the Emergency-notification and Driver-response Confirmation System(EDCS) for an autonomousvehicle safety that notifies the driver of an emergency situation and recognizes the driver's reaction in a situation where the driver is careless. The EDCS uses the emergency situation delivery module to make the emergency situation to text and transmits it to the driver by voice, and the driver response confirmation module recognizes the driver's reaction to the emergency situation and gives the driver permission Decide whether to pass. As a result of the experiment, the HMM of the emergency delivery module learned speech at 25% faster than RNN and 42.86% faster than LSTM. The Tacotron2 of the driver's response confirmation module converted text to speech about 20ms faster than deep voice and 50ms faster than deep mind. Therefore, the emergency notification and driver response confirmation system can efficiently learn the neural network model and check the driver's response in real time.