1
行政院國家科學委員會補助專題研究計畫成果報告
※※※※※※※※※※※※※※※※※※※※※※※※※※
※ ※
※
應用口哨於電腦音樂之搜尋與製作
※
※ ※
※※※※※※※※※※※※※※※※※※※※※※※※※※
計畫類別：□個別型計畫 □整合型計畫
計畫編號：NSC 90-2213-E-110-032
執行期間： 90 年 8 月 1 日 至 91 年 7 月 31 日
計畫主持人：李宗南
計畫參與人員： 許丕忠 沈鴻哲 潘進儒 陳雅筠
執行單位：國立中山大學資訊工程系
中 華 民 國 91 年 11 月 28 日
2
應用口哨於電腦音樂之搜尋與製作
Whistle for Musics – A Music Retrieval and Composition System
計畫編號: NSC NSC 90-2213-E-110-032
執行期間: 90 年 8 月 1 日 至 91 年 7 月 31 日
主持人：李宗南 教授 國立中山大學資訊工程系
一、中文摘要 在本計劃中，我們對口哨音樂與互動式 MIDI 音樂作曲器的設計作研究。基本上， 音樂的片段是由旋律、和絃，和節拍所組 成。在本系統中，首先將口哨轉換成音樂 的訊號。然後，這些音樂訊號被解析成數 種單音的旋律片段，藉由這些旋律片段， 系統中所提供的和絃及節拍可以自動地產 生。本篇研究以三方面描述口哨音樂。第 一，在電腦音樂系統中的互動式領域。第 二，將口哨的單音旋律轉換為 MIDI 訊號， 而這些轉換中的參數包含了定音、持續時 間，還有速度。最後，是討論音樂的表示， 以及將旋律調和為音樂的演算法。 關鍵字：口哨音樂Abstract
In this project, we study the design of an interactive MIDI music composer - “Whistle for Music”. Since a piece of music is composed of three parts: melody, chords, and rhythm. The system first accepts the whistling tune as the musical signals. Then those musicals signal are analyzed to derive many monophonic melodic phrases. Based on those melodic phrases, the supporting chords and drum rhythm patterns can be generated automatically.
This paper describes “Whistle for music” by three aspects. One is the interactive dimension for a computer music system. The other is monophonic melody transcription that transcribes the whistling input to MIDI notes with parameters in pitch, duration, and velocity. The last is algorithmic composition that needs musical representation and manipulation to harmonize a melody to
music.
Keywords: Whistle for Music, MIDI, Music
Retrieval, Query by Humming, Music Composition
二、緣由與目的
Most people assume that learning to play a musical instrument requires many years of practices in order to perform it well. Without a good command of any single musical instrument, one seldom tries to compose music. The research presented in this paper is aimed to reduce the barriers to produce computer music [1-3]. The proposed “Whistle for Music” is concerned with providing an easy and cost-effective way to enable average whistlers to compose and perform computer music. Imagining a person, who is not an expertise in any electronic music interfaces, needs a short piece of background music for a video game or a graphics demonstration. However, as a programmer, composing music is not his/her expertise. Therefore, providing an easy “tune writer” for the whistler who has only limited musical knowledge to compose a piece of computer music becomes an interesting task.
三、研究方法與成果
Figure 1 shows the system architecture of “whistle-to-music”. The authoring process of MIDI format music requires two-stage processes. The first stage is a transformative process called “melody transcription” that transcribes the whistling notes into melodic contour by analyzing the information of pitch, strength and duration. And the second stage is a
3
generative process called “algorithmic composition” that control regular rhythm for the transcribed melodic contour and adds harmonic accompaniment on this melody. The first stage of “whistle to melody” is real-time and transparent. The second stage is a generative one called “melody to music”.
The melody transcription component uses the real-time FFT algorithm to analyze the real performance of whistling input. The parametric controls for FFT are tempo and resolution, etc. The results of signal processing by FFT are displayed in the time-frequency map called spectrogram. They are the fundamental frequencies of whistled notes in the spectrogram. And those frequencies are mapped to the musical notes in the “MIDI grid”. MIDI grid is a graphic “piano roll” that can play MIDI notes.
Compositional knowledge is represented as contour operators that use user-controlled parameters to generate rhythm and harmony. Finally, musical synthesizer plays the transcribed melody and generated accompaniment together.
Figure 1. System architecture of “Whistle-to-Music”
In summary, “Whistle to Music” integrates both “melody transcription” and “computer-aided composition” to achieve an
interactive musical system. Melody transcription accepts the whistled tune to generate a melodic contour in the first step. Then by analyzing the monophonic melody, some intelligence (contour development) and knowledge of musical arrangement can be applied to create an interactive MIDI composer that responds in real-time to input from the whistler.
四. 計劃成果自評
Currently, we have developed an interactive
MIDI composer that contains three features: easiness, flexibility and generality. Some advantages are as follows:
(1). It enables the average whistlers or novices to try their musical potential easily.
(2). It establishes the prototype of MIDI music generation not only for the interface of microphone but also for other interfaces.
(3). It proposes a method for how to convert a simple melodic contour into a reasonable quality of musical composition.
五、參考文獻
1. Yim, S.; Ding,Y; George, E.B., “A real-time MIDI music synthesis system using sinusoidal modeling on a TI TMS320C32 digital signal processor, ”Multimedia Signal Processing, 1997., IEEE First Workshop on , 1997, Page(s): 469 –474.
2. Haus, G.; Sametti, A. “Scoresynth: a system for the synthesis of music scores based on Petri nets and a music algebra” Computer , Volume: 24 Issue: 7 , July 1991, Page(s): 56 –60.
3. Chih-Chin Liu, Jia-Lien Hsu and Arbee L. P. Chen. Efficient theme and non-trivial repeating pattern discovering in music databases. IEEE 1999.
