Abstract: A new method for approximate melody matching, called " linear alignment natching", is presented, and based on the new method, a query by humming system is implemented. Compared with the existing content-based music retrieval system, the new approach does not extend from approximate string matching, statistic model, and feature space methods. The new design is based on the geometrical similarity of melody contours, and both pitch and rhythm are taken into consideration. Experiments are conducted to test the new algorithm 62 humming clips are studied in a music database comprising 3864 songs, and LAM’s top-3 success rate is 90.3%, exceeding the traditional approximate string matching algorithm by 11% . This experiment result shows the validity of the "linear alignment matching" method and its feasibility of being used as the query engine in large digital music libraries.