Method for terahertz time-domain spectral analysis based on wavelet transform

The conventional terahertz spectral analysis based on Fourier transform needs to measure the terahertz reference signal without simples, and the terahertz spectra measurement is required in nitrogen. A novel method for terahertz time-domain spectral analysis based on wavelet transform is proposed to reduce terahertz spectral measurement steps. Firstly, the de-noising process about the spectroscopic data of the sample is achieved, and then the multi-scale wavelet-transform is realized by Coiflets wavelet base. The wavelet multi resolution entropies of samples under different wavelet decomposition scales are calculated, and the characteristic information describing samples is obtained directly. The experimental results indicate that the wavelet multi-resolution entropies, which are observably different between different samples but adequately stable between the same samples, can provide a valid quantitative criterion for the materials recognition. Spectral signal's wavelet multi-resolution entropies have good repetition in different humidity, and the average variance of the wavelet multi-resolution entropies is smaller than 0.05.