摘要
设计和实现了一个噪声环境下的哼唱信号处理系统。该系统首先利用减谱法对哼唱信号进行去噪处理,然后采用平均幅度差函数与自相关函数相结合的方法提取基频,再对其进行两次中值平滑处理,最后使用改进的扩窗法切分音符,并将其转化为音高和时间来描述旋律特征。实验结果表明,该系统具有良好的抗噪性,并能将哼唱信息准确有效地转化成旋律特征。在测试的60个歌曲片段中,各种唱法的平均正确率达到了91.4%。
A humming signal processing system in noise environment is presented.First,the spectrum subtraction method is used to eliminate the noise in the humming signals.Then,the pitch is extracted by combining the average magnitude difference function（AMDF） and the autocorrelation function（AUTOC）.Followed with two times median smoothing processing,the improved method of spread windowing is adopted to segment notes.The music melody are presented by the characteristics of pitch and time.The experiment′s results show that the humming signal can be converted to melodic characteristics exactly and effectively,besides with well antinoise performance.Among the 60 tested songs′ segments,the average right ratio comes up to 91.4%.
出处
《现代电子技术》
2010年第10期33-36,共4页
Modern Electronics Technique
关键词
哼唱检索
基音检测
音符切分
旋律描述
query by humming
pitch detection
notes segmentation
melody presentation
