數字簡譜系統與相關樂曲自動分析研究

Abstract

本論文研製之自動分析MIDI檔案產生數字簡譜系統，這個系統可將MIDI音訊檔轉譯為樂譜，並在MIDI資訊可能不完整的情況下(e.g.演奏即時錄製的MIDI檔案等)分析出正確的音符事件與整首樂曲的樂理性質，並且把音符訊息以相對簡單的方式呈現，使不太懂音樂理論的人也可以快速理解樂譜的表現。針對MIDI資訊記載不完整的檔案，透過多種分析與追蹤技術的運用，整合出一個簡單易懂的音樂呈現機制，不論是使用錄音介面所產生的MIDI檔案或從wav檔案轉譯的MIDI檔案，甚至是樂器演奏等資訊有所誤差的檔案能夠得到最佳化過的結果。 本系統整合了以下分析追蹤技術與規則理論： 1. 音名拼寫，此分析採用可即時拼寫音名的研究技術。 2. 調性分析，也是利用螺旋陣列研究取得最可能的調性。 3. 拍子追蹤與節奏的歸納，從權重周期的分析再歸納節拍。 4. 和絃的判斷。 5. 考慮音樂的通用性，我們加入了音樂理論的規則例如：數字音符標示法則、拍子的基本架構、調性轉移觀念、流行樂的調性標示法則和流行樂的和弦標示法則。

This study provides a number musical notation system which can be used to convert MIDI files to number musical notation. Writing music in number musical notation can be easily stored and transported electronically, which enables the understanding of music for with less music knowledge. The following analysis and tracking technologies and rule theories were integrated in number musical notation system: 1. Pitch spelling: An in-situ technology of pitch spelling was used. 2. Key finding: A spiral array model also was used to optimize the tonality of music. 3. Beats tracking and rhythm induction: This study adopted a process from the analysis of beats weights and duration to rhythm induction. 4. Chord identification: The center of effect in spiral array model was utilized to identify the chord of music. 5. Rule theories: Some rule theories were also integrated in this system, such as number musical notation rule, beat basic framework, tonality transfer, tonality and chord marking rule for pop music. iii Based on the integration of pitch spelling, tonality analysis, beats tracking and chord identification, number musical notation system can analyze the MIDI files without complete information, such as the files recorded from audio recorders or converted from WAV files, to find correct note events and the characteristics of music in MIDI files, hence get the optimized results.