Please use this identifier to cite or link to this item: http://hdl.handle.net/11455/18757
標題: 利用子母音之部分共同向量法於中文單音之辨識
Using Common Vector Approach to Recognize Isolated Mandarin Word Based on Vowel and Consonant
作者: 劉懿宸
Liu, Yi-Chen
關鍵字: 梅爾頻率倒頻譜係數;Mel-frequency cepstrum coefficient;特徵擷取;共同向量法;Feature Extraction;Common vector
出版社: 統計學研究所
引用: 1 Bayrakceken M.K., Cay M.A. and Barkana A., (2007),"Word Spotting Using Common Vector Approach", Signal Processing and Communications Applications, IEEE 15th , pp. 1 - 4. 2 Cevikalp H., Neamtu M., Barkana A., (2007),"The Kernel Common Vector Method: A Novel Nonlinear Subspace Classifier for Pattern Recognition", Systems, Man, and Cybernetics, Part B: Cybernetics, IEEE Trans. on Speech and Audio Processing, Vol. 37, Issue: 4 , pp. 937 - 951 3 Edizkan R., Gu ̈lmezog ̌lu M. B., Ergin S. and Barkana A., (2005),"Improvements on common vector approach for multi class problems", In 13th European Signal Processing Conference, Antalya, Turkiye. 4 Ergin S. and Gu ̈lmezog ̌lu M.B., (2007),"Face recognition based on face partitions using Common Vector Approach", Communications, Control and Signal Processing, ISCCSP 2008. 3rd International Symposium on, pp. 624-628. 5 Gu ̈lmezog ̌lu M. B., Dzhafarov V., Keskin M., and Barkana A., (1999),"A Novel Approach to Isolated Word Recognition", IEEE Trans. On Speech and Audio Processing, Vol. 7, no. 6, pp. 620-628. 6 Gu ̈lmezog ̌lu M. B., Dzhafarov V. and Barkana A., (2001),"The Common Vector Approach and Its Relation to Principal Component Analysis", IEEE Trans. On Speech and Audio Processing, Vol. 9, no. 6, pp. 655-662. 7 Fukunaga K. and Koonz W. L., (1970),"Application of the Karhuenn-Loeve expansion to feature selection and ordering", IEEE Trans. Comput., Vol. C-19, no. 4,pp. 311-318. 8 Iijima T., Genchi H., Mori K., (1973),"A theory of character recognition by pattern matching method", The Proc. of the 1st International Joint Conference on Pattern Recognition, Washington, DC, pp. 50–56. 9 Keser S. and Edizkan R., (2009)," Phonem-based isolated Turkish word recognition with subspace classifier", Signal Processing and Communications Applications Conference, IEEE 17th, pp. 93 - 96 10 Lakshmi C., Ponnavaikko M., Sundararajan M., (2009),"Improved Kernel Common Vector Method for Face Recognition", Machine Vision, ICMV ''09. Second International Conference on, pp. 13 - 17 11 Serkan G., Edizkan R., (2007),"Use of Novel Feature Extraction Technique with Subspace Classifiers for Speech Recognition", Pervasive Services, IEEE International Conference on, pp. 80 - 83 12 Watanabe S., Lambert P.F., Kulikowski C.A., Buxton J.L. and Walker R., (1967), "Evaluation and selection of variables in pattern recognition", in Compter and Information Sciences ii. New York: Academic, pp. 91. 13 Watanabe S. and pakvasa N., (1973),"Subspace method in pattern recognition", in Proc. 1st Int. Conf. Pattern Recog., Washington, DC, pp. 25-32. 14 陳漢良、李宗寶, (2012)。「利用共同向量法辨識中文單音及爆破氣音的探討」。碩 士論文,國立中興大學統計學研究所,台中。
摘要: 
本篇論文主要是利用共同向量法探討1391個中文單音之辨識。共同向量法是一個簡單且容易應用的方法,不只是可應用在語音辨識而且也可應用在臉部辨識。共同向量法是一種線性子空間分類器,在每一個類別裡,將所有訓練的特徵投影到一個唯一且共同的特徵上作為這個類別的語音模型。這篇論文中,我們運用梅爾頻率倒頻譜係數當作語音的特徵,將中文單音分割成子音與母音兩部分分別作共同向量法,觀察在不同的參數下,如子音母音的加權比重和建構語音模型所運用到的樣本點數的組合,去比較何種參數組合會有較好的辨識率。此次實驗是由十二位不同語者所錄製的語音資料庫去做辨識,權重設定為(子音,母音)=(0.1,0.9)到(0.9,0.1),取樣本點數為1到7。實驗結果發現,當權重設定為(0.5,0.5),取樣本點數為4時,最佳平均辨識率為82%。因此,共同向量法在中文單音辨識上有不錯的辨識率。

This paper is to investigate the 1391 monosyllable in speaker-dependent system. The method of common vector is used for the speech recognition. The method is simple and easy in application not only for speech recognition but also for face pattern recognition. The common vector approach is a linear subspace classifier. In each class, it projects all training features into a unique common feature as the model for the corresponding class. In this paper, We use the Mel-frequency cepstrum coefficient (Mfcc) as the feature in the recognition. The mandarin monosyllable divided into consonants and vowels two part, in which the common vector approach is contructed as model, respectively. The different weights are then given for each consonant and vowel parts as the parameters in speech recognition. In the work, the set of training samples will also be considered as the parameter in recognition. From the exprimental results, we find that when the weight equals to (0.5,0.5) for consonant and vowel parts, and the training samples is 4, the best speech recognition will be up to 82%.
URI: http://hdl.handle.net/11455/18757
其他識別: U0005-0907201317195100
Appears in Collections:統計學研究所

Files in This Item:
File SizeFormat Existing users please Login
nchu-102-7100018006-1.pdf886.76 kBAdobe PDFThis file is only available in the university internal network    Request a copy
Show full item record
 

Google ScholarTM

Check


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.