基于DTW算法的音乐音符特征识别  

Musical Note Feature Recognition Based on DTW Algorithm

作　　者：高菲[1] GAO Fei(School of Music,Huainan Normal University,Anhui 232001,China)

机构地区：[1]淮南师范学院音乐学院

出　　处：《新乡学院学报》2019年第9期50-53,共4页Journal of Xinxiang University

摘　　要：使用传统方法对音乐音符特征分类的精度较低,对音乐音符特征识别准确性较差,为此提出一种基于DTW算法的音乐音符特征识别方法。通过对音乐音符进行分析,将音符标准距离下的相似度矩阵作为DTW距离的音乐音符相似度矩阵,根据相似度矩阵给出音符特征选择标准,将它作为音乐音符特征子集优化目标函数。通过多种群的运行状态,从多个种群中,获取较高的音乐音符分类精度的期望值,将其作为多种群分类质量评价值,对每个音乐音符作出全局客观评价,根据评价值进行音乐音符分类,完成识别。实验结果表明,所提方法的音乐音符特征分类精度较高、音乐音符特征识别准确性好且整体识别耗时较少。The classification accuracy of musical note features in traditional methods is low, which leads to poor recognition accuracy of musical note features.Thus, a musical note feature recognition method based on DTW algorithm was proposed in this paper.Through the analysis of musical notes, the similarity matrix under the standard distance of musical notes was used as the musical note similarity matrix of DTW distance.According to the similarity matrix of musical notes, the selection criteria of note features were given.It was regarded as a subset of musical note features to optimize the objective function.Through the running state of multi-population, the expected value of higher musical note classification accuracy was obtained from multiple populations, which was regarded as the quality evaluation value of multi-population classification to make global objective evaluation of each musical note, to carry out musical note classification and complete the recognition.The experiment results showed that the proposed method had higher classification accuracy, better recognition accuracy and less time consuming.

关 键 词：DTW算法 音乐音符 特征识别 期望值 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]