基于鲁棒音阶特征和测度学习SVM的音乐和弦识别

发布时间：2018-04-14 19:09

本文选题：和弦识别 + 音阶轮廓特征　；参考：《信号处理》2017年07期

【摘要】：和弦识别是音乐信息检索领域重要的研究内容之一,在信息处理、音乐结构分析以及推荐系统等方面具有重要的作用。为了降低人声对和弦进程的影响且恢复和弦所对应的谐波信息,文章分别对频谱中和弦所对应的谐波信息和人声信息进行建模,构建双目标优化问题,对和弦所对应的谐波信息进行有效重建,同时去除人声;其次,对谐波信息进行降维处理得到鲁棒性的音阶轮廓特征;最后为了提高支持向量机性能,文章采用测度学习的方法得到马氏距离,并使用马氏距离替换支持向量机的高斯核函数的欧氏距离,使得支持向量机的判别函数包含有数据的空间分布信息。最终实验结果表明,同基于现今流行的和弦识别算法相比,提出的和弦识别算法识别正确率提高3.5%~12.2%。
[Abstract]:Chord recognition is one of the important research contents in the field of music information retrieval. It plays an important role in information processing, music structure analysis and recommendation system.In order to reduce the influence of the human voice on the chord process and restore the harmonic information corresponding to the chord, the harmonic information and the human-voice information of the chord in the spectrum are modeled in this paper, and the two-objective optimization problem is constructed.The harmonic information corresponding to the chord is reconstructed effectively, and the human voice is removed. Secondly, the harmonic information is reduced to obtain robust scale contour features. Finally, in order to improve the performance of support vector machine,In this paper, the Markov distance is obtained by the measure learning method, and the Euclidean distance of the Gao Si kernel function of the support vector machine is replaced by the Markov distance, so that the discriminant function of the support vector machine contains the spatial distribution information of the data.The experimental results show that the recognition accuracy of the proposed algorithm is higher than that of the popular chord recognition algorithm by 3.5% and 12.2%.
【作者单位】：天津大学电子信息工程学院;
【基金】：国家自然科学基金资助项目(60802049,61471263) 天津市自然科学基金重点项目(16JCZDJC31100)
【分类号】：TN912.3;TP391.3

【相似文献】