鋼琴音樂的韻律與調性之自動轉譜系統

Automatic Transcription of Rhythm and Tonality in Piano Music

吳冠勳(Guan-Xun Wu)

指導教授 ： 張文輝

國立交通大學/電機學院/電信工程系所/碩士(2018年)

摘要

隨著電子音樂科技的蓬勃發展，以樂器數位介面(MIDI)音樂為基礎的電子鋼琴已日漸普及化。節奏與旋律決定了音樂的曲風，也是演奏者對於樂曲理解與詮釋的重要依據。但MIDI音樂沒有記錄演奏內容的音值與調性資訊，因而無法有效評估彈奏者的音樂表現力。為了提升鋼琴初學者的自主練習成效，本論文旨在開發一個基於MIDI音樂的韻律與調性之自動轉譜系統。前人研究面臨的挑戰是音符的持續時間取決於拍速及音值，但人為彈奏的諸多不穩定因素影響其音值及樂曲調性的自動判定。我們提出兩階段處理的音值識別機制，利用基於卡爾曼濾波器的拍速追蹤器，配合按鍵起始時間差計算其拍速不變性的特徵，再透過隱藏式馬可夫模型的建構求出彈奏樂曲之音值序列。在調性分析的系統製作上，主要是利用基於五度圈而建構的螺旋陣列模型，配合音高以及持續時間所計算的音符影響中心，進而自動判別彈奏樂曲之調性。實驗結果證實，我們提出的兩階段處理音值識別機制，可大幅提升其音符時值及樂曲調性自動判別的準確性。

關鍵字

樂器數位介面 ； 音樂轉譜 ； 音值識別 ； 卡爾曼濾波器 ； 拍速追蹤 ； 隱藏式馬可夫模型 ； 螺旋陣列模型 ； 調性分析

並列摘要

Music transcription is one of the most fundamental and challenging problems in music information processing. The difficulty arises from the fact that performed note lengths and local tempo fluctuate from the nominal note lengths and score-indicated tempo. This study presents a statistical method for use in music transcription that can estimate note value and key from monophonic MIDI performance signals. The system implementation is divided into three modules: tempo tracking, note value recognition and key finding. The first step is to apply the Kalman filtering algorithm to track the time-varying tempo in performed music. After that, the inter-onset intervals (IOIs) are normalized by local tempo to obtain tempo-invariant features. The note value recognition is formulated as a problem to find the most likely sequence of intended note value by using hidden Markov model (HMM). Finally, we apply the spiral array model to estimate the key by using both the recognized note values and pitches of performed notes. Experimental results show that the proposed two-step procedure significantly improves the performances in note value recognition as well as key finding.

並列關鍵字

MIDI ； music transcription ； note value recognition ； Kalman filtering ； tempo tracking ； hidden Markov model ； spiral array model ； key fidning

參考文獻

[1] R.B. Dannenberg, M. Sanchez, and A. Joseph, "A computer-based multi-media tutor for beginning piano students," Interface-Journal of New Music Research, 19(2-3), 1990, pp.155-173

[2] R.B. Dannenberg, M. Sanchez, A. Joseph, R. Joseph, R. Saul, and P. Capell, "Results from the piano tutor project," in Proc. of the 4th Biennial Arts and Technology Symposium, March 1993.

[3] T. Kitamura and M. Miura, "Constructing a support system for self-learning playing the piano at the beginning stage," in Proc. of International Conference on Music Perception and Cognition, 2006.

[4] G. Percival, Y. Wang, and G. Tzanetakis, "Effective use of multimedia for computer-assisted musical instrument tutoring," in Proc. of EMME'07, Sep. 2007.

[5] 謝苑玫,電腦輔助音樂教學研究-音符與休止符認知教學之成效分析,國立台 灣師範大學音樂研究所碩士論文,1987.