流行音樂風格與情感反應研究

隨流行音樂產業的蓬勃發展，音樂設計展演使用科學研究方法，可以更有效的製作出符合演出情境效果的演唱會。本研究透過八大曲風的音樂樣本蒐集、情感反應尺度篩選、音樂訊號分析、支援向量機與音樂特徵分析等研究步驟，可以了解音樂構成的元素，以及其展現於樂曲當中的特色與情感反應的關聯性，建構適合的流行音樂演唱會演出曲目，可以讓觀眾融入並沉浸於特定的演唱會音樂風格與情境當中。本研究首先篩選出代表性流行音樂風格曲目，而後進行音樂樣本收集與感性辭彙篩選，透過因素分析篩選分類標籤及使用支援向量機建立情感分類模型。研究結果顯示，情感反應尺度結果歸類出三組分類標籤，分別是「特殊性」、「感受性」及「變化性」，並針對這三組分類標籤分別進行音訊分析，可以看出各分類標籤與音量、音色這兩項特徵之間的關係。最後透過支援向量機所獲得的最佳訓練模型平均準確率為93.3%，證明了本研究結果的可行性，分析結果也提供具體而明確的音樂設計建議，並作為未來延續音樂與情感反應研究的重要參考資料。

關鍵字

流行音樂風格；音樂特徵；音樂情感反應；因素分析；支援向量機

並列摘要

As the popular music industry thrives, scientific methods have become a more effective way of designing music concerts with befitting performance effects. This study collected samples of eight major music styles, selected affective response dimensions, analyzed music signals, employed a support vector machine, and analyzed musical characteristics to understand the composing elements of music and the connection between the features they display in music and affective responses. Constructing a suitable popular music concert repertoire can immerse the audience in certain concert music styles and scenarios. We first chose songs representative of various popular music styles and then collected music samples and selected descriptive adjectives. Using factor analysis, we selected category labels and used a support vector machine to establish an affective categorization model. The study results revealed three category labels for the affective response dimensions: uniqueness, feeling, and variation. We conducted signal analysis based on these three category labels to derive the relationships between these category labels and the features volume and timbre. The mean accuracy rate of the optimized and trained model obtained using a support vector machine was 93.3%, thereby demonstrating the feasibility of the study results. The analysis results also provide clear and specific suggestions for music design and serve as important reference for future research on music and affective responses.