摘要
改进的BPM算法首先将基于上下文的节拍周期估计方法和节拍追踪算法结合起来,再通过起始点探测算法得到的音符频次,利用音符和节拍的关系细分节拍探测区间,接着通过三种置信度的计量方式来评估每个探测区间的置信度,得到BPM值及其相关特征.实验结果得出改进的BPM音频节奏特征提取算法在特征提取的准确率、音乐情感分类效果方面比传统算法都更具有优越性.
The improved BPM algorithm combines the estimation method of the context-based beat cycle with the beat tracking algorithm firstly,obtaining the frequency of the notes through the starting point detection algorithm,subdividing the beat detection interval by the relationship between notes and the rhythms,and then evaluating the confidence of each detection interval by three confidence measurement formulas.The BPM value and its related features are obtained accordingly.The experimental results show that the feature extraction algorithm of the improved BPM audio rhythm is superior to the traditional algorithm in terms of the accuracy of feature extraction and the effect of music emotion classification.
作者
吴昊
吴畏
孙晓燕
WU Hao;WU Wei;SUN Xiao-yan(School of Computer Science,Hefei Normal University,Hefei 230601,China;School of Software Engineering,XI＇AN Jiaotong University,Xi＇an 710049,China;Department of information and Control Engineering,China University of Mining and Technology,Xuzhou 221116,Jiangsu,China)
出处
《兰州文理学院学报（自然科学版）》
2018年第4期52-59,共8页
Journal of Lanzhou University of Arts and Science(Natural Sciences)
基金
国家自然科学基金项目“定性指标偏好感知进化优化在个性化搜索中的应用”(61473298)
安徽省自然科学基金项目“基于自适应粒子滤波算法的目标跟踪计算模型及方法研究”(1708085QF157)
安徽省高校优秀青年人才支撑计划项目“基于仿生学和graph cut理论的粒子滤波跟踪算法研究”(gxyq2017050)
