A Study on Implementation of Voice Control System Using DSP Chip 蔡學承、李立民
E-mail: 386773@mail.dyu.edu.tw
ABSTRACT
In this research, we design and implement a voice control system and use it to control a robot. The voice control system is
implemented on a TMS320C6713 DSK board, which is a digital signal processor circuit board from Texas Instruments. The speech feature vector consists of 12 linear prediction derived cepstral coefficients, log energy and their first and second order derivatives.
The automatic speech recognition system used in this study is based on the hidden Markov model. The Matlab and Simulink are used as a high level system development environment in this study. The system designed as a Simulink model consists of several interconnected modules, each of which is either represented as a block of embedded function or a built-in block in Simulink. We use the Embedded Coder and the Real-time Workshop to convert the Simulink model the associated Matlab programs into C language programs. Using the Code Composer development environment, the C language programs are compiled and linked into an executable program and then downloaded into the DSK board. The recognition result is then used to drive an infrared remote controller to send a control signal to control the robot. The parameters of the speech models are obtained from offline training. The modules of the real-time automatic recognition system include a voice acquisition and digitization unit, a framing unit, an end-point detection unit, a feature extraction unit, a data buffering unit, a core speech decoding unit, and a control signal output unit.
Experimental results show that the voice control system can effectively control the robot.
Keywords : digital signal processor、automatic speech recognition、hidden Markov Model、voice control Table of Contents
封面內頁 簽名頁 中文摘要．．．．．．．．．．．．．．．．．．．．．．iii 英文摘要．．．．．．．．．．．．．．
．．．．．．．．iv 誌謝．．．．．．．．．．．．．．．．．．．．．．．．v 目錄．．．．．．．．．．．．．．．
．．．．．．．．．vi 圖目錄．．．．．．．．．．．．．．．．．．．．．．．viii 表目錄．．．．．．．．．．．．
．．．．．．．．．．．x 第一章 緒論．．．．．．．．．．．．．．．．．．．．．1 1.1 研究緒論．．．．．．．．．
．．．．．．．1 1.2 研究背景．．．．．．．．．．．．．．．．2 1.3 研究目的．．．．．．．．．．．．．．．．3 第 二章 語音特徵參數擷取．．．．．．．．．．．．．．．4 2.1 語音特徵參數擷取．．．．．．．．．．．．4 2.2 梅爾頻 率倒頻譜參數．．．．．．．．．．．14 第三章 隱藏式馬可夫模型．．．．．．．．．．．．．．17 3.1 以距離及動態規 劃為觀點的語音辨識．．．．17 3.2 隱藏式馬可夫模型的組成．．．．．．．．．20 3.3 使用Viterbi方法做語音模型訓練 與語音辨認．22 3.4 使用EM演算法做模型訓練．．．．．．．．24 3.4.1 正算程序．．．．．．．．．．．．．25 3.4.2 逆算程序．．．．．．．．．．．．．27 3.4.3 模型參數重估．．．．．．．．．．．28 第四章 實驗過程．．．．．．．
．．．．．．．．．．．32 4.1 語音資料庫格式．．．．．．．．．．．．．32 4.2 狀態數與維度的差異．．．．．．．
．．．．33 4.2.1 維度比較．．．．．．．．．．．．．35 4.3 實際測試．．．．．．．．．．．．．．．．40 4.4 紅外線 遙控器原理．．．．．．．．．．．．43 第五章 結論．．．．．．．．．．．．．．．．．．．．46 5.1 結論．．．．．
．．．．．．．．．．．．．46 5.2 未來研究方向．．．．．．．．．．．．．．47 參考文獻．．．．．．．．．．．．
．．．．．．．．．．48 REFERENCES
[1] J. Makhoul, “Linear prediction: A tutorial review,” Proc. of IEEE, vol. 63, no. 4, pp. 561-580, Apr. 1975.
[2] 王小川, “語音訊號處理”, 全華圖書, 台北縣土城市, 2008.
[3] 維基百科, http://en.wikipedia.org/wiki/Hidden_Markov_model.
[4] K.-F. Lee, “Automatic Speech Recognition: The Development of the SPHINX System,” Kluwer Academic Publishers, Boston, 1989.
[5] B. Bogert, M. Healy and J. Tukey, “The quefrency analysis of time series for echoes,”Proc. Symp. On Time series Analysis , New York , J.
Wiley , 1963 [6] S. Davis, P. Mermelstein, “Comparing of Parametric Representations for Monosyllable Word Recognition in Continuously Spoken Sentence,” IEEE Trans. on Acoustic, Speech and Signal Processing, 1980.
[7] 蔡博宇, “以FPGA為平台之嵌入式類神經網路語音辨識系統實現”, 碩士論文, 電機工程學系, 高雄大學, 2009 [8] 李胥瑜, “結合基因 演算法及經驗模態分解進行強健性語音辨識與FPGA晶片實現”, 碩士論文, 資訊工程學系, 高雄大學, 2009 [9] 楊俊哲, “隱藏式馬可夫模
型之語音辨識在電視控制系統之應用”, 碩士論文, 工程科學及海洋工程學研究所, 台灣大學, 2012 [10] 蔡宜亨, “具有強健性語音辨識的 無線語音控制系統研製”, 碩士論文, 資訊工程學系, 高雄大學, 2011 [11] TEXAS INSTRUMENTS Technical Reference,
http://c6000.spectrumdigital.com/dsk6713/V2/docs/dsk6713_TechRef.pdf [12] HOLTAK data sheet, http://www.holtek.com.tw/pdf/consumer/6221_2v170.pdf
