Abstract:

An improved pitch recognition algorithm based on the relevant processing and Fast Fourier Transform（FFT） is proposed.The music signal is preprocessed by the use of the three-level center clipping function.Then the pitch is estimated by using the related processing method.A filter is designed by using the parameters which are aroused from the estimated pitch value.After filtering the music signal the precise frequency result can be got by using FFT method.The algorithm has higher recognition accuracy than the traditional time domain processing method，and more effectively recognizes the pitch which is rich of harmonic than the traditional harmonic peaks method，and has less amount of computing than the wavelet algorithm.The experimental simulation proves that the method is feasible，fast and reliable.

Key words: recognition of pitch, fundamental tone detection, Fast Fourier Transform（FFT）, endpoint detection

摘要： 提出了一种基于自相关处理和快速傅里叶变换（FFT）的改进的单音音高识别算法。利用修正的三电平中心削波函数对音乐信号进行预处理，再自相关处理估计基音周期，以估计周期为参数设计滤波器，音乐信号滤波后用FFT实现频域的准确定位。该算法比传统的时域处理法具有更高的识别精度，能比谐波峰值法更有效地解决谐波丰富、基频分量小的信号的识别，且运算量比小波算法小。经实验仿真验证该方法可行、快速可靠。

关键词: 音高识别, 基音检测, 快速傅里叶变换, 端点检测