计算机与现代化

• 应用与开发 • 上一篇    

基于权值分配的音符识别研究

  

  1. 南京理工大学自动化学院,江苏南京210094
  • 收稿日期:2016-08-11 出版日期:2017-03-29 发布日期:2017-03-30
  • 作者简介:卞毓伟(1992-),男,江苏盐城人,南京理工大学自动化学院硕士研究生,研究方向:系统工程,模式识别; 郭玲(1976-),女,辽宁沈阳人,副教授,博士,研究方向:多媒体信号处理。

Research on Notes Recognition Based on Weight Distribution

  1. School of Automation, Nanjing University of Science & Technology, Nanjing 210094, China
  • Received:2016-08-11 Online:2017-03-29 Published:2017-03-30

摘要: 在基于音高的音乐识别中,基音周期是音符识别中最重要的参数之一。本文以钢琴乐曲为例,提出一种基于权值分配的基音检测算法。在基音周期自相关检测的基础上,分析音乐信号的帧样本,并利用基于权值分配的改进方法检测音符基频,能够有效提高基频识别精度。

关键词: 基音周期, 端点检测, 自相关, 帧样本, 权值分配

Abstract: Pitch is one of the most important parameters of music signal in the pitchbased music recognition. In this paper, a new pitch detection algorithm based on weight distribution is proposed by taking piano music as an example, in which autocorrelationbased pitch detection is carried out first, then each sample frame of the music signal is analyzed, and finally an improved weight distribution method is used to detect the fundamental frequency of the music. Experiments show its effectiveness in improving the recognition accuracy.

Key words: pitch period, endpoint detection, autocorrelation, frame sample, weight distribution

中图分类号: