基于音乐认知原理的音乐旋律发现技术

发布时间：2018-02-26 03:01

本文关键词： 音乐认知旋律发现听觉显著度常数Q变换旋律流　出处：《信号处理》2010年10期 　论文类型：期刊论文

【摘要】：旋律是音乐主题思想的最重要表现手段,分析音乐的旋律、研究智能旋律分析处理方法是音乐信息检索领域的重要课题之一。本文根据脑神经科学及认知心理学关于人类对音乐感知特点的研究成果,引入听觉显著度(AS)的概念,提出了基于音乐认知理论的模拟人类认知过程的旋律发现技术。在前期处理阶段,针对音乐特性采用常数Q变换(CQT)建立音乐的频谱统计模型,采用贝叶斯理论计算每个半音子带数据分布的听觉显著度特征,利用时序神经网络检测各个时刻的听觉变化,得到旋律分量的候选。在后期处理阶段,我们提出了表达形式接近乐理与认知的旋律流(MelodyStream)的概念,以人对音乐和弦感知结果作为先验知识,进行旋律候选分量的规范化处理。在包含各种乐曲风格的实验音乐数据库上,验证了所提取结果同人类听感的接近程度,根据旋律流来捕捉传统旋律线获得了75%的准确率,主观听感打分对旋律流的接受度超过90%。
[Abstract]:Melody is the most important means of expression of music theme thought. The research of intelligent melodic analysis and processing is one of the important subjects in the field of music information retrieval. According to the research results of brain neuroscience and cognitive psychology on the characteristics of human's perception of music, the concept of auditory saliency is introduced in this paper. This paper presents a melody discovery technique based on music cognition theory to simulate human cognitive process. In the early stage of processing, a spectrum statistical model of music is established by using constant Q transform (CQT) to analyze the characteristics of music. Bayesian theory is used to calculate the auditory saliency characteristics of each semitone subband data distribution, and the temporal neural network is used to detect the auditory changes at each time, and the melody component candidates are obtained. We put forward the concept of Melody Stream, in which the expression form is close to musical theory and cognition. We take the result of people's perception of music chords as a priori knowledge to normalize melody candidate components. In the experimental music database, which contains all kinds of musical styles, The accuracy rate of capturing the traditional melody line according to the melody flow is 75% and the acceptance degree of the subjective auditory sense to the melody flow is more than 90.
【作者单位】：哈尔滨工业大学计算机科学与技术学院;
【基金】：国家自然科学基金项目No.60772076 语言语音教育部-微软重点实验室开放基金资助项目(HIT.KLOF.2009015) 黑龙江省留学归国基金LC03C10的支持
【分类号】：J614.6

【相似文献】