Music Signal Segmentation Using Tree-Structured Filter Banks

For music signal segmentation, an 11-stage tree-structured filter bank that leads to wavelet packet analysis was designed and implemented. The tree-structured filter bank has sufficient frequency resolution over a wide range of frequencies, while the time resolution at high frequencies satisfies the minimum time resolution for the human ear. To improve selectivity, a new family of quadrature mirror filters (QMFs) with a narrow transition region and a low reconstruction error was implemented.