http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
- 中文 을 입력하시려면 zhongwen을 입력하시고 space를누르시면됩니다.
- 北京 을 입력하시려면 beijing을 입력하시고 space를 누르시면 됩니다.
RISS 인기검색어
- 검색키워드
- 저자 : 고시영 (검색결과 37 건)
- 무료
- 기관 내 무료
- 유료
-
-
高時暎 慶一大學校 1987 論文集 Vol.3 No.1
A system for V-UV-S detection of Korean speech is developed to improve the defections of Energy-spectra method. This detection system is based on variation of the frequency distribution of Energy-spectra and Zero crossing rates. The experimets show that the UV-V and S-UV miss-detections are corrected contently, for application to Korean speech recoqnition especialy.
-
다양한 조건에 따른 TDMA와 로터리 버스 중재방식의 성능비교
고시영,Koh, Si-Young 한국정보통신학회 2012 한국정보통신학회논문지 Vol.16 No.9
일반적인 버스 시스템 구조는 공용버스 내에 여러 개의 마스터와 슬레이브, 아비터 그리고 디코더 등으로 구성되어 있다. 복수의 마스터가 동시간대에 버스를 이용할 수 없으므로, 아비터는 이를 중재하는 역할을 수행한다. 아비터가 어떠한 중재방식을 선택하는가에 따라 버스 사용의 효율성이 결정된다. 기존의 중재 방식에는 Fixed Priority방식, Round-Robin 방식, TDMA 방식, 로터리 방식 등이 연구되고 있다. 본 논문에서는 최근에 개발된 TDMA 방식과 로터리 방식에 대하여 성능을 비교해 보고, 성능개선을 위한 방법을 제안하려고 한다. The general bus system architecture consists of masters, slaves, arbiter, decoder and so on in shared bus. As several masters can't use a bus concurrently, arbiter plays an role in bus arbitration. In compliance with the selection of arbitration method, The efficiency of bus usage can be determined. Fixed Priority, Round-Robin, TDMA, Lottery arbitration are studied in conventional arbitration method. In this paper, we draw the performance comparison of TDMA and Lottery bus arbitration policy developed recently due to farious conditions and propose the methods of performance improvement.
-
高時暎 慶一大學校 1988 論文集 Vol.4 No.1
Thepaper proposes new method for classifying signal frame of Korean speech into voiced,unvoiced silence frame using Energy-spectra and ZCR merhod. In additions, the unvoiced is classified into fricative and stop consonant by the lenth of the sound segment. The proposed system is relatively simple but itsperformence is fairly accurate and reliablr.
-
Switched-Capacitor 低域通過여파기의 設計에 關한 硏究
高時暎 慶一大學校 1987 論文集 Vol.3 No.1
A study on the designe of SC LPF The two 4-th order active low pass filters are designed using the SC-FDNR method and the biguad SC block which used bilinear Z-transformation method respectively. The experiments show that the frequency characteristics of the both SCF are similar to that of the active RC and passive filter and it is negligible phase delay of the both SCF compared to the active RC and passive filter. The SCF which fabricated in IC could be expected much improvement of performances.
-
高時暎 慶一大學校 1999 論文集 Vol.16 No.2
In this paper, performance comparison of 4-Algorithms, based CHMM, in the Korean isolated Digits Recognition is tested. From the experimented Results, the recognition rates of CHMM+△CEP and CHMM+MIX are creased (1.13%) and (1.79%), respectively, then the case of CHMM(recognition rate : 86.27%), but the rate of NNPHMM decreased (7.93%) that is lowest rate however, we will continue to experiment to find the optimal algorithem for NNPHMM whitch is becoming one or the most effective methods.
-
神經網에 의한 韓國語 音聲의 單音 및 音表 Segmentation
高時暎 慶一大學校 1991 論文集 Vol.7 No.1
In this paper, the segment algorithem for speech signal using neural network proposed and investigated resonability by computer simulation. Propased neural network is organized multi-layer perceptrons with bias. Matching rate experiment is carried out of continuous vowel, place names and 4 conneted Digits. As a results, the case of continuous vowel is obtained 100%, the case of place names is 99% and the case of Digits is 97%.
-
高時暎 慶一大學校 1995 論文集 Vol.12 No.2
In this paper, co-articulation one-stage dynamic programming algorithm combined with Hidden Marcov Model is proposed for redognizing Korean 4 connected digits. The training process is used the continuous distribution HMM using the Baum-welch algorithm. Average recognition rate is 91.7 percent for 35 kinds of 4 connected digits uttered by 5 male speakers. In comparision with the values without the process of HMM, these value are upgraded by about 1.9 percent.
-
高時暎 慶一大學校 1997 論文集 Vol.14 No.2
This paper studies on the recognition for 25 sentences of robot control. The performances of the proposed system using one pass DP method with the context control, conbineded with distribution HMM, are investigated and compared with O(n) DP method without the context control. The recognition rates of the one pass DP method and O(n) DP method are 98.6% and 90.6% respectively. This result reveals remarkable improvement in the context control
-
高時映 慶一大學校 1999 論文集 Vol.16 No.5
In this paper, The Recognition system, using FSN0OSDP combined whth CHMM and CMHMM respectively, is proposed for recogmizing korean control commands and tested From the experimented Results, the average recognition rates of the CHMM and CMHMM turn out to be 96.5% and 97.2% respectovely This results show that CMHMM is better method than CHMM and the proposed method can be development to practical recognition ssystem for the limited vocabulary of control commands.
연관 검색어 추천
이 검색어로 많이 본 자료
활용도 높은 자료
