一种基于小波变换的太赫兹时域光谱分析方法

Method for Terahertz Time-Domain Spectral Analysis Based on Wavelet Transform

传统太赫兹光谱傅里叶分析法要求必须先测量出无样品时的THz信号作为参考信号,对测量环境要求高,提出了一种基于小波变换的太赫兹时域光谱信号处理方法,以简化太赫兹光谱测量的步骤。对空气中直接测得的样品光谱信号进行去噪预处理,然后利用Coiflets正交小波基进行多尺度下的小波分解,进而计算出各样品不同尺度下的小波多分辨信息熵,得到表征样品的特征信息。实验结果表明,在太赫兹频段,不同样品的小波多分辨信息熵值差异显著,同一样品则保持稳定;不同湿度下光谱信号的小波熵值重复性好,平均偏差小于0.05。

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.