【正文】 JIU JIANG UNIVERSITY 畢 業(yè) 論 文（設(shè) 計(jì)） 題 目 基于聲紋的說(shuō)話人特征識(shí)別 英文題目 Speaker feature recognition based on the voiceprint 院 系 專(zhuān) 業(yè) 姓 名 年 級(jí) 指導(dǎo)教師 20xx 年 6 月 九江學(xué)院學(xué)士學(xué)位論文 I 摘 要 說(shuō)話人識(shí)別 是一項(xiàng)根據(jù)語(yǔ)音波形中反映說(shuō)話人生理和行為特征的語(yǔ)音參數(shù)，而 自動(dòng)識(shí)別說(shuō)話人身份的技術(shù) 。它也常被人們稱(chēng)為聲紋識(shí)別技術(shù)，是生物認(rèn)證技術(shù)的一種，其基本思想就是運(yùn)用某種匹配方法進(jìn)行特征識(shí)別，從而確定說(shuō)話人的身份。 目前已知的語(yǔ)音特征包括 基音周期、語(yǔ)譜圖 、 自相關(guān)系數(shù)、能量、平均幅度、過(guò)零率、 共振峰、線譜對(duì)、線性預(yù)測(cè)系數(shù) （ LPC） 、線性預(yù)測(cè)倒譜（ LPCC）、Mel頻率 倒譜（ MFCC）等。 本文介紹了 說(shuō)話人識(shí)別的概念、原理及其識(shí)別實(shí)現(xiàn)的方法，指出了說(shuō)話人識(shí)別技術(shù)的應(yīng)用前景。通過(guò)在 、線性預(yù)測(cè)倒譜和Mel頻率倒譜等 特征參 數(shù)進(jìn)行提取、分析、對(duì)比、識(shí)別實(shí)現(xiàn)一個(gè)簡(jiǎn)單的說(shuō)話人識(shí)別系統(tǒng)，實(shí)驗(yàn)結(jié)果表明實(shí)驗(yàn)正確、有效。 關(guān)鍵字： 說(shuō)話人識(shí)別；特征參數(shù)； 基音周期 ； 線性預(yù)測(cè)倒譜 ； Mel頻率 倒譜 基于聲紋的說(shuō)話人特征識(shí)別 II Speaker feature recognition based on the voiceprint Abstract Speaker recognition is the voice parameters in a speech waveform which reflects the speaker39。s physiological and behavioral characteristics, and automatic identification technology to speaker identity. It is also often referred to as the voiceprint recognition technology, a biometric authentication basic idea is to use a matching method for feature recognition, in order to determine the identity of the speaker. Currently known voice features include pitch, spectrogram, since the correlation coefficient, energy, average magnitude, the zero crossing rate, formant, the line spectrum of the Linear Prediction Coefficient (LPC), Linear Prediction Cepstrum (LPCC) , Mel Frequency Cepstral (MFCC). This article describes the speaker identification concepts, principles and implementation methods of identification, and pointed out the prospect of speaker recognition technology. By the platform, voice pitch, linear prediction cepstrum and Mel Frequency inverted spectra characteristic parameter extraction, analysis, contrast, identify a simple speaker recognition system, experimental results show that the experiment is correct, effective . Key Words： Speaker Recognition； Feature Parameter； Pitch； Linear Prediction Cepstral Coefficient； Mel Frequency Cepstral Coefficient 九江學(xué)院學(xué)士學(xué)位論文 III 目 錄 摘 要 ........................................................................................................................... I Abstract........................................................................................................................II 目 錄 ........................................................................................................................ III 引 言 .......................................................................................................................... 1 第一章 說(shuō)話人識(shí)別研究 .............................................................................................. 3 說(shuō)話人識(shí)別研究的意義 .................................................................................. 3 說(shuō)話人識(shí)別應(yīng)用領(lǐng)域 ...................................................................................... 3 說(shuō)話人識(shí)別的技術(shù)優(yōu)勢(shì) .................................................................................. 4 說(shuō)話人識(shí)別研究的難點(diǎn)和熱點(diǎn) ...................................................................... 5 說(shuō)話人識(shí)別技術(shù)研究的難點(diǎn) ............................................................... 5 說(shuō)話人識(shí)別研究的熱點(diǎn) ....................................................................... 7 影響說(shuō)話人識(shí)別性能的因素 .......................................................................... 7 論文的內(nèi)容安排 .............................................................................................. 9 第二章 說(shuō)話人識(shí)別的基本介紹 .............................................................................. 10 語(yǔ)音的基礎(chǔ)知識(shí) ............................................................................................ 10 語(yǔ)音的產(chǎn)生原理 ................................................................................. 10 語(yǔ)音產(chǎn)生模型 ..................................................................................... 10 語(yǔ)音信號(hào)的預(yù)處理技術(shù) ..................................................................... 12 說(shuō)話人識(shí)別的分類(lèi) ........................................................................................ 14 說(shuō)話人識(shí)別的基本原理 ................................................................................ 16 說(shuō)話人識(shí)別的常用特征 ................................................................................ 18 說(shuō)話人識(shí)別系統(tǒng)的結(jié)構(gòu)框架 ........................................................................ 18 說(shuō)話人識(shí)別的主要模型 ................................................................................ 20 說(shuō)話人識(shí)別系統(tǒng)評(píng)價(jià)標(biāo)準(zhǔn) ............................................................................ 22 第三章 特征參數(shù)的提取 ............................................................................................ 24 倒譜 ............................................................................................................... 24 同態(tài)處理基本原理 ............................................................................ 24 復(fù)倒譜和倒譜 .................................................................................... 25 線性預(yù)測(cè)倒譜 (LPCC)的提取 ....................................................................... 25 LPCC 的介紹 ................................................................................... 26 LPCC 的提取過(guò)程 ........................................................................... 27 Matlab 中實(shí)現(xiàn) LPCC 的提取 ......................................................... 27 Mel頻率倒譜 (MFCC)的提取 ...................................................................... 28 基于聲紋的說(shuō)話人特征識(shí)別 IV Mel 頻率介紹 .................................................................................. 28 MFCC 提取過(guò)程 ............................................................................. 29 Matlab 中實(shí)現(xiàn) MFC