基於低音聲部特徵萃取的和弦行進與風格演算法之音樂情緒分類研究

Abstract

近年來伴隨著MP3此種以超高壓縮比與超小失真程度音樂格式的出現，以及網路快速的成長，帶動了數位音樂流行的風潮。從古至今的各種音樂型態，會由於樂曲編排、樂器編制、樂手風格、及地區表現手法等，產生出許多不同的音樂情緒。 市面上眾多的播放程式也設計了多樣化的歌曲分類方式，例如演唱(奏)者、專輯名稱、年份、音樂風格等分類法，用以幫助使用者整理歌曲清單，但卻鮮少看到一種針對「音樂情緒」的分類方式。因此本文則利用音樂中BASS此項樂器為基礎，利用數位訊號處理(DSP)的技術，根據BASS的音符使用，計算其音樂的和弦行進，大小調的運算，音樂風格的區別，來製作一圖像式的音樂情緒反應介面，讓使用者對該歌曲的內容有初步的了解。

Numerous style of music produces very different listening mood and emotion from song arrangement, musical instrument element, musician playing style, district feature etc. Many music players have designed the diverse song classification way, such as singer or performer, album name, years or style, it is music classification ways, and various metadata need to be created for each music piece. The content-based music information retrieval is under aggressive research. This study used MP3file, proposed a classification in music emotion based on the instrument “bass guitar”. The developed bass line feature extraction calculated the chord progression, music tonal, music genre, and established a real-time graphic user interface on music emotion while playing. It may help user realize the music they listening in time.