http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
PARCOR 係數 抽出 Algorithm實現에 관한 硏究
裵澈秀 관동대학교 1988 關大論文集 Vol.16 No.2
The purpose of this paper is to compare Durbin, Le-Roux, Le-Roux lattice algorithms with which PARCOR extract coefficients from original speech. PARCOR coefficients are extracted according to each algorithms. In this study characteristics of PARCOR coefficients are estimated on comparing the original speech with the synthesized speech using the PARCOR coefficient extract. Le-Roux method is superior to Durbin in computation quantity, easier to control in hardware of real time speech processing. We cannot get the LPC coefficients by means of using Le Roux method. Le Roux lattice method is the most useful algorithms to extract both PARCOR coefficients and LPC coefficients which are required in recognition.
시간축정규화와 WEIGHTING 계수를 포함한 화자인식에 관한 연구
裵澈秀 관동대학교 1989 關大論文集 Vol.17 No.1
Preliminary results of using orthogonal parameters from PARCOR coefficients for speaker recognition problem are presented. The recognition process include standard distance measure for eigenvalues, log likelihood ratio distance measure using DTW for optimal path, and euclidean distance measure for orthogonal parameters and weighted orthogonal parameters, in order to consider dynamic and weighted features of speech signals. To generate the database for experiments, seven speakers were asked to utter the same sentences ten times. Using the MKM(modified K-means method), we extracted the reference pattern from the data obtained from four sessions. Experimental results using euclidean distance measure for orthogonal parameters and for weighted orthogonal parameters according to optimal path, showed 85.7% and 91.4% of recognition rate, respectively.
裵撤秀 관동대학교 1993 關大論文集 Vol.21 No.2
본 논문에서는 음성인식 뉴럴 네트워크를 위한 뉴럴퍼지학습법에 관하여 연구하였다. 본 논문에서 연구한 학습방법에서 교사신호에 하나의 확정된 어휘정보를 주는 종래방법과는 달리 비확정적(FUZZY) 어휘 정보를 주며 뉴럴퍼지학습법은 역전파법으로 실현된다. 뉴럴 퍼지학습법의 유효성을 확인하기 위해 28개 DDD지역명을 대상으로 인식·확인 하기위해 실험을 하였으며, 그결과 91.9%의 향상된 인식을 얻을수 있었다.
마이크로 프로세서에 依한 디지탈 濾波器의 實現에 關한 硏究
裵澈秀 관동대학교 1985 關大論文集 Vol.13 No.3
This paper aims to realize the digital filter using Microprocessor. The use of distributed arithmetic is known to be able to provide increased speeds by avoiaing time-consuming multiplications. High component efficiency is also achieved if the in herent control capabi-lities of the microprocessor are fully utilised, in particular where support horware element are used. By way of examples, Butter-worth and chebyshev low-pass fillter were impleme-nted by use of microprcessor and their characteristics were examined and analyzed.