Design And FPGA Implementation Of Wavelet Filter For Spectral Signal

With the continuous development of spectral analysis technology,spectral measurement has been widely used in industry,mining and daily life.When the spectrometer is used for spectral analysis of substances,the spectral processing data are easily affected by noise and interference,which reduces the accuracy of the analysis results.Therefore,it is urgent and practical to Denoise the spectral data.Although the traditional denoising methods such as Fourier analysis,curve fitting and smoothing are effective in the field of data processing,they have little effect in the face of complex spectral data.As a new means of signal processing,wavelet analysis has become a natural tool for spectral signal analysis and processing because of its flexible thinning ability in time and frequency domain.In this paper,the structure of the spectrometer,the noise source analysis of spectral signal and the content of spectral signal preprocessing are introduced in detail.For the measured noisy spectral signals,we design a wavelet threshold denoising algorithm based on wavelet theory.By comparing the combinations of different wavelet basis functions,wavelet orders,layers and thresholds,the denoising results obtained from many experiments are analyzed,and the best combination of denoising effect of traditional methods is determined from the point of view of signal-to-noise ratio,root mean square difference and smoothness.On this basis,considering the limitation of soft and hard threshold,this paper proposes a semi-soft threshold function and weighted threshold function to obtain better denoising effect.In addition,based on the concept of Bayesian threshold in two-dimensional image processing,we improve the traditional wavelet threshold to form an improved wavelet Bayesian threshold filtering algorithm.compared with the traditional wavelet threshold algorithm,a better denoising effect is obtained.Then,based on the study of wavelet threshold filtering algorithm,we choose the improved Bayesian hard threshold scheme combined with simulation tools to implement the wavelet filter on the FPGA hardware platform.First of all,based on the idea of data framing,we use the symmetrical structure of double multipliers to realize the hardware structure of the filter.Then each module is modeled separately and its function is verified by Verilog description language simulation.Finally,the results of hardware calculation are imported into MATLAB.By comparing the filtering results obtained by two different methods,it is proved that the data processing result of the wavelet threshold filter realized by hardware is consistent with that of the software algorithm.that is to say,the filtered spectral signal can meet the requirements of actual detection and analysis.Finally,through the analysis of the running speed and occupied resources of the wavelet filter,it is concluded that the wavelet filter based on FPGA has absolute advantages in realtime processing of spectral data.In addition,this paper makes some experiments and researches on the wavelet de-noising scheme in the spectral field.These practical works have positive significance for the theoretical research and development of wavelet de-noising of spectral signals.