基于嵌入式系统的实时声音频谱分析技术
发布时间：2018-12-10 13:43
【摘要】：声压频谱分析是考虑人耳对不同频率成分的声音的感受的不同,进而通过傅里叶变换等获得其准确频谱特性的技术。声压频谱分析是后续声学分析的基础,同时在声学测量,噪声污染,健康医疗,降噪减噪,故障诊断,国防建设等中都具有重要的应用。本课题从理论分析到硬件及算法设计介绍了基于嵌入式系统的手持式声压分析平台的实现过程。提出了一套精度较高,运算量较小,实时性较好,可操作性较强的手持式声学频谱分析方案。 本课题对基于嵌入式系统的声压频率计权、频谱分析进行了深入的分析与研究,首先介绍了基于卷积的频率计权实现及基于傅里叶变换、快速傅里叶变换的声压频谱分析方法。考虑卷积运算的运算量比较大,本文提出了基于重叠相加DFT/IDFT的频率计权优化算法,并分析了两者的算法复杂度及最优帧长选择标准。针对Matlab库函数设计的频率计权滤波器与本课题目标计权滤波器频响相差较大,介绍了一种基于DFT/IDFT的频率计权滤波器的设计方法。考虑FFT变换是基于复数域的,研究了声压信号的FFT/IFFT加速实现算法。最后对本课题的实验平台的实现进行了系统分析,介绍了主要的硬件构成和相应的驱动程序开发,分析了输入输出通道的校准方法以及在快速傅里叶变换过程中需要使用到旋转因子,位反系数和分离因子的计算方法。 对本课题的研究方案,分别在Matlab上进行仿真分析和嵌入式平台上进行实际测试,结果表明,基于重叠相加DFT变换的频率计权实现方法与基于卷积的实现方法结果一致,而且随着滤波器系数的增大前者性能快速提升,对计权后的信号进行时域积分,证明本课题的研究方法的计算结果满足GB3240-1982一级标准。对正弦信号进行频谱分析,也可以得到近似脉冲信号的主瓣,较小的边瓣和过渡。
[Abstract]:The sound pressure spectrum analysis is a technique that takes into account the different perception of the human ear to the sound with different frequency components, and then obtains the exact spectrum characteristics by Fourier transform and so on. Sound pressure spectrum analysis is the basis of subsequent acoustic analysis, and it has important applications in acoustic measurement, noise pollution, health care, noise reduction, fault diagnosis, national defense construction and so on. From theoretical analysis to hardware and algorithm design, this paper introduces the realization process of handheld sound pressure analysis platform based on embedded system. In this paper, a handheld acoustic spectrum analysis scheme with high precision, low computation, good real-time and high maneuverability is proposed. In this paper, the sound pressure frequency measurement and frequency spectrum analysis based on embedded system are deeply analyzed and studied. Firstly, the realization of frequency weighting based on convolution and the method of sound pressure spectrum analysis based on Fourier transform and fast Fourier transform are introduced. Considering the complexity of convolution operation, a frequency weight optimization algorithm based on overlapping additive DFT/IDFT is proposed, and the complexity of the two algorithms and the optimal frame length selection criteria are analyzed. Aiming at the difference between the frequency response of the frequency weighting filter designed by Matlab library function and that of the target weighted filter, a design method of the frequency weighted filter based on DFT/IDFT is introduced. Considering that FFT transform is based on complex domain, the FFT/IFFT acceleration algorithm of sound pressure signal is studied. Finally, the realization of the experiment platform is analyzed systematically, and the main hardware structure and the corresponding driver development are introduced. The calibration method of input and output channels and the calculation method of rotation factor, bit inversion coefficient and separation factor are analyzed. The results show that the frequency weighting method based on overlapping plus DFT transform is consistent with that based on convolution. Moreover, with the increase of filter coefficient, the performance of the former is improved rapidly, and the time domain integral of the weighted signal is carried out, which proves that the calculation results of this research method meet the GB3240-1982 first order standard. The main lobe, small sidelobe and transition of the approximate pulse signal can also be obtained by spectrum analysis of the sinusoidal signal.
【学位授予单位】：浙江大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TB526
[Abstract]:The sound pressure spectrum analysis is a technique that takes into account the different perception of the human ear to the sound with different frequency components, and then obtains the exact spectrum characteristics by Fourier transform and so on. Sound pressure spectrum analysis is the basis of subsequent acoustic analysis, and it has important applications in acoustic measurement, noise pollution, health care, noise reduction, fault diagnosis, national defense construction and so on. From theoretical analysis to hardware and algorithm design, this paper introduces the realization process of handheld sound pressure analysis platform based on embedded system. In this paper, a handheld acoustic spectrum analysis scheme with high precision, low computation, good real-time and high maneuverability is proposed. In this paper, the sound pressure frequency measurement and frequency spectrum analysis based on embedded system are deeply analyzed and studied. Firstly, the realization of frequency weighting based on convolution and the method of sound pressure spectrum analysis based on Fourier transform and fast Fourier transform are introduced. Considering the complexity of convolution operation, a frequency weight optimization algorithm based on overlapping additive DFT/IDFT is proposed, and the complexity of the two algorithms and the optimal frame length selection criteria are analyzed. Aiming at the difference between the frequency response of the frequency weighting filter designed by Matlab library function and that of the target weighted filter, a design method of the frequency weighted filter based on DFT/IDFT is introduced. Considering that FFT transform is based on complex domain, the FFT/IFFT acceleration algorithm of sound pressure signal is studied. Finally, the realization of the experiment platform is analyzed systematically, and the main hardware structure and the corresponding driver development are introduced. The calibration method of input and output channels and the calculation method of rotation factor, bit inversion coefficient and separation factor are analyzed. The results show that the frequency weighting method based on overlapping plus DFT transform is consistent with that based on convolution. Moreover, with the increase of filter coefficient, the performance of the former is improved rapidly, and the time domain integral of the weighted signal is carried out, which proves that the calculation results of this research method meet the GB3240-1982 first order standard. The main lobe, small sidelobe and transition of the approximate pulse signal can also be obtained by spectrum analysis of the sinusoidal signal.
【学位授予单位】：浙江大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TB526
【参考文献】
相关期刊论文 前10条
1 宋毅s，
本文编号：2370673
本文链接：https://www.wllwen.com/guanlilunwen/gongchengguanli/2370673.html
