钢琴音乐信号的检测技术研究

发布时间：2018-02-02 02:46

本文关键词： 钢琴音乐信号时频分析起始点检测基频检测多基频检测　出处：《南京理工大学》2017年硕士论文　论文类型：学位论文

【摘要】：随着人们的生活水平日益提高,越来越多的人走进音乐课堂,通过学习一门乐器来提升气质和修养。然而初学者在学习钢琴时需要花费大量时间在纠错上,这使得学生的课堂学习效率低下,教师的辅导工作繁重。本文对钢琴音乐信号的检测技术进行研究,旨在对钢琴演奏音乐信号特征分析,结合先验乐谱信息,对音乐信号进行音符检测和识别。主要工作如下:1)研究了时频分析技术在钢琴音乐信号中的应用。仿真分析了基于短时傅里叶变换、连续小波变换和Hilbert-Huang变换的音乐信号时频图,重点讨论了窗函数的形状和窗长对STFT时频分析结果的影响,并进一步讨论了二进小波分解和EMD分解的方法,为音乐检测技术的研究打下了基础。2)讨论了音符起始点检测。以音乐信号的短时能量、高频内容(HFC)、短时频带方差和频谱差值等参数为信号特征,结合单音音乐和多音音乐仿真分析了基于不同信号特征音符起始点检测性能。仿真结果表明HFC更有利于钢琴音乐的音符起始点检测。3)讨论了单音音乐的基频检测方法。针对短时自相关、小波分析和谐波峰值法等单基频检测算法只适用于一定频率范围的局限,提出了一种基于频谱峰值排序的检测方法。仿真表明频谱峰值排序法可以实现全频带的基频检测。4)针对音乐信号的多基频检测,基于信号分解的思想,提出了基于子带分解与FFT联合检测多基频的方法。首先将频谱峰值排序法延伸到多基频检测领域,实现了高、低两通道处理;其次针对现阶段小波变换只能检测单基频的局限,探讨了二进小波变换和FFT联合的检测方法,实现了在二进制频带下的多基频检测;然后根据EMD自适应分解的特性完成了 EMD与FFT联合检测。仿真结果表明EMD与FFT联合能实现多基频的精确检测。最后结合乐谱完成了整首乐曲的检测识别,检测结果表明本文的钢琴音乐信号的检测方法具有较好的检测效果。
[Abstract]:With the increasing improvement of people's living standards, more and more people enter the music classroom to improve their temperament and accomplishment by learning a musical instrument. However, beginners spend a lot of time on correcting mistakes when learning the piano. This makes students' classroom learning inefficient and teachers' guidance work arduous. This paper studies the detection technology of piano music signal aiming at analyzing the characteristics of piano playing music signal and combining the information of a transcendental music score. The main work is as follows: 1) the application of time-frequency analysis technology in piano music signal is studied. The simulation analysis is based on short-time Fourier transform. The time-frequency diagram of music signal based on continuous wavelet transform and Hilbert-Huang transform is discussed. The influence of window shape and window length on STFT time-frequency analysis results is discussed. Furthermore, the methods of dyadic wavelet decomposition and EMD decomposition are discussed, which lays a foundation for the study of music detection technology. The high frequency content of HFC is characterized by some parameters, such as the variance of short time frequency band and the difference of frequency spectrum. The performance of detecting the starting point of notes based on different signal features is analyzed by combining the simulation of single tone music and multi-tone music, and the simulation results show that HFC is more advantageous to the detection of the starting point of piano music. 3). This paper discusses the fundamental frequency detection method of single tone music, aiming at short time autocorrelation. Wavelet analysis and harmonic peak detection algorithm are only suitable for the limitation of a certain frequency range. A detection method based on spectrum peak sorting is proposed. The simulation results show that the spectrum peak sorting method can realize the fundamental frequency detection of the whole frequency band. 4) for the music signal multi-fundamental frequency detection, based on the idea of signal decomposition. In this paper, a method based on sub-band decomposition and FFT is proposed to detect multi-fundamental frequency. Firstly, the method of spectrum peak ranking is extended to the field of multi-fundamental frequency detection, which realizes high and low two-channel processing. Secondly, aiming at the limitation that wavelet transform can only detect single fundamental frequency at present, the combined detection method of binary wavelet transform and FFT is discussed, and the multi-fundamental frequency detection in binary frequency band is realized. Then, according to the characteristic of EMD adaptive decomposition, EMD and FFT joint detection. The simulation results show that the combination of EMD and FFT can achieve the accurate detection of multi-fundamental frequency. Finally, the detection and recognition of the whole piece of music is completed in combination with music score. The results show that the detection method of piano music signal in this paper has better detection effect.
【学位授予单位】：南京理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：J624.1

【相似文献】