基于比值算法和能量重心算法的语音信号频率估计及对比研究

Frequency estimation and comparison of speech signal based on ratio algorithm and energy center of gravity algorithm

近年来,信号处理技术在生活中网络通信、电子图像、医疗、工业等领域广泛应用。但是对无限长的信号进行测量和运算时,截取其中的有限信号片段进行周期延拓处理,得到的虚拟的无限长信号经过傅立叶变换等数学处理会出现频谱泄露现象。为了提高语音信号频率估计的精度,本文使用比值法和能量重心法分别对加入噪音前后的理想正弦信号测试其校正效果,接着基于能量重心算法,测试加入汉宁窗前后的信号频率估计效果,最后选用两种更贴近生活中的语音信号的复杂信号验证了实验结论。仿真结果表明,基于能量重心算法实现的频率估计结果准确度更高;基于某一固定频率校正算法,加入汉宁窗后的频率估计结果准确度更高。本文第二部分介绍了比值校正法、能量重心校正法的公式及公式推导原理;第三部分介绍了本研究的四个实验的步骤并以数据和图像的方式展示了实验结果;第四部分总结了实验工作,并对未来进一步研究进行展望。

Nowadays,signal processing technology has been widely used in many fields in our daily lives.However,when dealing with an infinite-length signal by Fourier transform,it often leads to spectrum leakage.To improve the accuracy of signal frequency estimation,this paper tested the accuracy of Ratio method and Energy Centro-baric method by using the ideal sinusoidal signal before and after adding noise,based on Energy Centro-baric method,estimated the accuracy of estimating signals’frequencies before and after adding Hanning window,and finally chose two complex signals which are closer to the signals in life to validate the conclusion.The result showed that Energy Centro-baric method was more accurate;based on a fixed correction algorithm,the frequency estimation with Hanning window was more accurate.The second part of this paper introduces the formula of ratio correction method and energy center of gravity correction method and their derivation principle.The third part introduces the four experimental steps of this study and presents the experimental results in the form of data and images.The fourth part summarizes the experimental work and looks forward to the future research.