音乐信号稀疏分析方法研究

A Novel Sparse Analysis Method for Music Signals

利用稀疏分解思想对音乐的类型、体裁、调式进行分析的方法在音乐的量化分析中占有重要地位,并已得到了广泛的应用.在传统的音乐稀疏化分析方法中,较少考虑系数矩阵内稀疏性能与音乐类别的联系,为了解决这一问题,我们设计了一个由稀疏分解结果获得的稀疏性能评价指标对音乐样本稀疏质量和重构效果进行了度量,并通过实验验证了该指标与重构误差及音乐体裁、调式存在明显的相关性,结果未来可以用于音乐类型分析工作的可靠参照.

The idea of using sparse decomposition to analyze music genre,pitch and melody has been widely used in the quantitative analysis of music.In traditional music sparse analysis,the relationship between the sparsity of the coefficient matrix and music genre is usually ignored.We designed a performance metric on the sparsity of the sparse representation matrix,so the sparse and reconstruction performance of music samples can be evaluated.The experiment result shows that this metric was correlated to the music genre,melody and a reliable performance index on music analysis.