Real-time Acoustic Spectrum Analysis Technology Based On An Embedded System

Audio spectrum analysis is a kind of technology to obtain to accurate spectrum characteristics by the technique of Fourier Transform after taking the fact that huma’s perception is different while the components of frequency is different. Audio spectrum analysis is the basis for the subsequent analysis of the acoustic. It is widely used in acoustic measurements, noise pollution, health care, noise reduction, fault diagnosis, national defense construction, etc. In this paper, we introduce the implementation of handheld audio spectrum analyzer detailly from the aspects of theoretical analysis, hardware and algorithm.Present a solution with high precision, faster speed, real-time, easy to use.In this artical, we conduct a deep-in research on audio spectrum analysis.Firstly, present a method of frequency weighting based on convolution and means for spectrum analysis based on Fourier transform, fast Fourier transform. Taking the fact that convolution will consume a large amount of computation into account, propose to optimize the process by the overlap-add DFT/IDFT algorithms. Analysis the complexity of both algorithm, obtain the selection rule of frame length.Since the frequency response is quite different with the target if we design the frequency weight filter with Matlab library, a novel way based on DFT/IDFT was adopted.To improve the performace,combine two real into a complex to accelerate the calculation of DFT/IDFT. Generally analyze the realization of experiment system, the hardware construction and Linux dirve development.Finally, algorithm of calibration for input and output channels as well as the calculation of rotary factor, inverse factor, split coefficient for DFT/IDFT are described detailly.Simulation with MATLAB simulink toolbox and experiment on embedded system were implemented later. The result shows that both kinds of methods get the same result and satisfy the GB3240-1982. The result of spectrum analysis for Sinusoidal signal is pulse signal with smaller side lobes and transition zone which is exactly what we need.