歐陽明Ouhyoung, Ming臺灣大學:資訊工程學研究所張元治Chang, Yuan-ChinYuan-ChinChang2010-06-022018-07-052010-06-022018-07-052008U0001-2401200815595400http://ntur.lib.ntu.edu.tw//handle/246246/184863電腦科技成長迅速的時代，各大公司競爭推出多媒體系統的加值服務已經在市面上普遍所見，結合不斷創新與獨領潮流的多媒體加值服務亦成為業者追求的目標。觀察近期市面上較新的多媒體應用，大多是將以往各平台上的多媒體服務轉移到個人電腦上，其中包括將點唱系統搬上個人電腦，並且透過網路更新推出的MTV，使得使用者不再侷限於購買昂貴且更新不方便的傳統點唱機或是外出至KTV消費。 在本論文中具有一項多媒體加值服務。由於市面上具備點唱功能的機器，升降音調都需要手動調整，對於多數未受過音樂訓練的大眾，調整音調後容易發現不一定適合自己的音域，因此本系統推出具備自動升降音調的功能，將演唱時所錄下的聲音與原唱歌手的聲音做比對，找出音高的差距，並將配樂移調至符合演唱者的音域，使演唱者可以按照習慣的演唱方式，輕鬆的演唱。論文中自動升降音調的功能由音高偵測與升降音調兩部份組成。傳統上音高偵測快速簡便的方式是在時域上實作，但考慮到複音偵測的準確度，在頻域上做偵測可以得到較佳的結果。實驗結果顯示，靠近原本音高的偵測準確度相當良好，而音高相差愈多則準確度會慢慢下降。第二部份-升降音調的功能，傳統作法主要亦分為時域與頻域，但經過實際比較之後，考慮到播放的音質以及升降音調的容忍範圍，採用頻域上Phase Vocoder的作法。透過Phase Vocoder此作法移調的範圍在上下一個八度得到的音質都相當良好，對於常用的移調範圍約為上下5個半音則相當足夠。Value-added services on multimedia system are becoming popular. The latest multimedia applications are transferring from different platforms to personal computers, including the Karaoke system. Moreover, newest MTV downloaded from internet relieves users from buying expensive and difficult-to-update Karaoke machines and from spending much money in KTV.An value-added service is proposed in the thesis, namely, an automatic key shifting system for Karaoke systems on PC. On all Karaoke platforms, we have to shift keys personally. Most untrained people do not know how to shift keys properly. Hence, we build the automatic key shifting system. Recording the voice and analyzing the pitch, we can shift the pitch of the accompany music to fit users’ gamut.In this thesis, the automatic pitch shifting function is made of two components: pitch tracking and pitch shifting. In general, pitch detectiion is performed efficiently in time domain. However, for this research, for better accuracy, pitch detectiion is implemented in frequency domain. itch shifting can also be carried out in time domain or frequency domain. After considering the acoustics effects and the tolerances of pitch shifting range, we use PV (Phase Vocoder) in frequency domain. We can get a better acoustics performance after shifting pitch within an octave. Besides, the shifting range is sufficient for general usage.口試委員會審定書i謝ii文摘要iii文摘要iv錄v表目錄viii圖目錄ix一章 序論1.1研究動機與目的1.2相關研究概述1.3問題描述1.4論文架構2二章 音高辨識方法3.1時域音高辨識方法32.1.1 ZC(Zero Crossing)32.1.2 ACF(Auto Correlation)52.1.3 AMDF(Average Magnitude Difference Function)62.1.4其它時域辨識方法72.1.5時域音高辨識總結7.2頻域音高辨識方法82.2.1Hamming框化8.2.2DFT92.2.3 CQ(Constant Q Transform)11.2.4其他頻域辨識方法13.2.5頻域音高辨識總結13.3 音高辨識比較14.4 其他音高辨識方法14.5 本章總結14三章 音高移調方法16.1 時域音高移調方法163.1.1 PSOLA (Pitch Synchronous Overlap Add)163.1.2 Resample183.1.3 Wsola (Waveform Similarity overlap-add19.2 頻域音高移調方法203.2.1 Phase Vocoder20.3 本章總結24四章 音樂自動移調系統254.1 系統流程254.2 參數設定264.3 程式實作304.4 本章總結35五章 實作測試結果討論365.1樣本比較36 5.1.1單音比較36 5.1.2透過Phase Vocoder升降調後比較38 5.1.3 受試者錄音樣本比較445.2 受試者問卷525.3執行速度535.4 本章總結55六章 結論57考文獻58application/pdf3883425 bytesapplication/pdfen-US音高辨識音高偵測自動移調移調方法Pitchpitch scalingpitch shiftingpitch trackingpitch detectingphase vocodermodekaraokethesishttp://ntur.lib.ntu.edu.tw/bitstream/246246/184863/1/ntu-97-J94922026-1.pdf