基于Praat软件的哼唱检索中参数选取分析

Characteristic Parameters Analysis in Query by Humming Based on Praat Software

摘要: 哼唱检索技术是一种基于内容的音乐检索技术。在哼唱检索技术中，声音特征参数的选取至关重要，它直接影响检索的最终结果。文章通过Praat语音分析软件，对声音的四种特征参数(能量、共振峰、MFCC-Mel倒谱系数、基音)分别进行了不同的计算机模拟实验，最终通过比对选取出符合哼唱检索要求的特征参数。

Abstract: Humming retrieval technology is a content-based music retrieval. In humming retrieval, it is es- sential to select the sound characteristic parameters, which will directly affect the final result of the search. Through Praat speech analysis software, different computer simulation experiments were carried out respectively to analyze the four characteristic parameters (energy, formants, MFCC-Mel Cepstral, pitch) of sound. By contrasting the results of these experiments, those charac- teristic parameters that can meet the humming retrieval request were selected.