基于动态分割和加权综合匹配的音乐检索算法

MIR Approach Based on Dynamic Thresholds Segmentation and Weighted Synthesis Match

在模式识别领域中,音乐检索算法由于其易行和高效的特性而得到越来越多的重视。但是音符分割不精确和匹配模版不统一等问题直接影响MIR算法精确度。为解决上述问题,提出基于动态阈值分割和加权综合匹配算法。动态设定幅差步长,根据变化阈值分割音符,以提高音符分割的准确性。采用标准乐谱频率,转换音高模板实现输入归一化,以增强匹配的精确性。融合直接匹配和DTW匹配,以加强算法适应性和鲁棒性。最后由实验证明该方法的有效性。

In the pattern recognition field, music information retrieval （MIR） is receiving more and more attention. To improve the accuracy of MIR system, a novel automatic algorithm based on dynamic thresholds note segmentation and weighted synthesis match is presented. Dynamic steps are set for amplitude difference, get dynamic thresholds to segment notes and improve segmentation veracity. Music scores are transformed into fiequency templates by standard frequency and the input signal pitch is normalized to enhance the accuracy. To improve the adaptability and the robustness, traditional approximate string match algorithm and DTW algorithm are synthesized. The simulation results are provided to show the validity of the algorithm.