摘要

Query-by-Humming involves retrieving music with a melody that matches the hummed query. An improved Query-by-Humming system for extracting pitch contour information based on a fuzzy inference model is introduced. In addition, an improved content-based music repeating pattern extraction model is introduced. Our bar-indexing method can extract the melody, identify repeating patterns and handle polyphonic MIDI files. To verify the effectiveness of the system, 15 volunteers recorded queries that were fed as input to the system and the longest common subsequence (LCS) was used to identify the most related top N matches. The system achieves 70% accuracy among the top 5 items retrieved.

  • 出版日期2015-8