Research On Signal Recovery Method For Removal Of Water Vapor Effects In THz-TDS

Terahertz time-domain spectroscopy technology is one of the hot spots in the research of terahertzapplication currently.THz-TDS plays a very important role in the application of terahertz spectral imaging and sample optical parameters measurement.The resonance effect of water vapor molecules in the air on THz signal has a huge impact on the measurement of terahertz signal in the THz-TDS experiment.In recent years,many scholars have studied the method of removing the influence of water vapor molecules on the THz signal.Based on the research of resonance absorption characteristics of water vapor molecules in the THz band,this paperachieved the recovery of THz signal through Wiener filter and BP neural network.By focusing on the basic principle and algorithm of multiscale wavelet decomposition and reconstruction,this paper designs a new multiscale signal recoverymethod which combines the Wienerfilter,BP neural network and Waveletmulti-scale decomposition.The damaged THz signal is decomposed into the components of different scales and processed by Wiener filtering and BP neural network to reconstruct the signal of the damaged THz signal,which leads to the realization of the signal recovery.The main tasks in this paper include:(1)Study and analyze the absorption characteristics of water vapor molecules.Design signal recovery scheme according to THz signal characteristics and water vapor responding characteristics.Signal recovery system block diagram based on multi-scale is designed.(2)Complete the construction and calibration of THz-TDS system.THz signal is collected and processed.By calculating the dynamic range of terahertz time-domain spectroscopy system and the system noise ratio,the research range of THz signal is selected.(3)Multi scale decomposition of THz signals based on wavelet multi-scale analysis method.In order to select the standard of the SNR of the wavelet hard threshold de-noising,the optimal wavelet bases are selected from the Daubechies wavelet clusters and the Symlet wavelet families with different filter lengths.According to the frequency characteristics of all levels of wavelet component decomposition,the decomposition series is determined,and the time-frequency characteristics of THz signal at different levels are compared and analyzed.(4)According to the recovered effect of regulating the beta parameters of the filter property,transferred to the optimal Wiener filter design and realize the simulation of Wiener filterof THz signal recovery results.The BP neural network is constructed and the learning and training of the neural network is completed,and the results of THz signal recovery by BP neural network are simulated.Recovered effect based on spectral contrast and absorption curve of recovered signals.(5)The recovery of THz signal is completed by the method of multiscale signal restoration designed in this paper,and the results are simulated and the recovery effect is analyzed according to the spectrum and the absorption curve.The two parameters-fluctuation rate and the relative lifting capacity are used as two important standards to judge the effects of THz signal recovery.The result shows that the multi scale signal recovery method has the best effect.In the research,lots of damaged THz signal are selected to be recovered through the four different recovery methods:Wiener filter,BP neural network,Wiener filter combined with BP neural network and the multiscale signal recoverymethod designed in this paper.Through selectingthe eight sets of experimental results to make the comparison,conclusion of the experiment and the further analysis is completed.