改进阈值提升小波和自适应滤波器的开放光路红外光谱去噪

大气污染物的主要组成成分为挥发性有机物(VOCs)，傅里叶变换红外光谱技术(FTIR)是现阶段应用广泛的挥发性有机物在线测量方法。开放光路获取到的大气红外光谱(OP-FTIR)易受各种噪声污染，如何有效、快速的去除红外光谱中的噪声是大气在线实时监测系统研究的热点。综合利用提升小波变换结构简单、运算量低的优点以及最小均方误差自适应滤波器的自动调节参数以达最优化滤波的性能，提出了一种改进阈值提升小波结合自适应滤波的红外光谱去噪算法。该算法先通过改进阈值小波系数的提升小波去噪，在去噪的同时保留更多光谱特征信息，然后使用提升小波变换分解出的高频系数重构出噪声相关信号，将其作为最小均方误差自适应滤波器的参考输入进行二次滤波处理，最终获得的去噪信号很好的去除了与特征光谱频谱重叠的噪声信号。分别对人工添加噪声的标准红外光谱和合肥市市区上空实测开放光路红外光谱进行去噪处理，结果显示使用该算法处理后的光谱信噪比(SNR)较离散小波传统阈值去噪方法高出3db，均方根误差(RSME)平均减少30%左右，运行时间减少46%。表明该算法计算简单、运行速度快，对于大气环境监测实时消噪系统具有重要的实际应用意义。

Open-Path Fourier Transform Infrared Spectrum De-Noising Based on Improved Threshold Lifting Wavelet Transform and Adaptive Filter

The major component of atmospheric pollutants is volatile organic compounds (VOCs), Fourier transform infrared spectroscopy (FTIR) is a widely used VOCs on-line measurement method at the present stage. The infrared spectrum obtained by open path (OP-FTIR) is easily to be polluted by various noise. Therefore, the development of effective and rapid methods to remove the noise in infrared spectrum is a crucial problem in the research of on-line atmospheric real-time monitoring system. The lifting wavelet transform (LWT) has the advantages of simple structure, and low computation; the least mean square algorithm (LMS) adaptive filter has the performance of automatically adjusts parameters to achieve optimal filtering. From above algorithm performance we proposes a infrared spectroscopy denoising algorithm combined with improved threshold LWT denoising and LMS adaptive filter. The algorithm first uses improved threshold LWT denoising preserve more spectral information and then uses the LWT decomposition of the high-frequency coefficients to reconstruct the noise correlation signal. Take this noise as the reference input of the LMS adaptive filter, the final denoising signal is effective for the removal of the noise signal overlap with spectral spectrum. In the experimental part, the standard infrared spectrum plus noise and the measured infrared spectrum of open optical channel over Hefei city were denoised respectively, the results show that the signal-to-noise ratio of the spectral signal processed by the proposed algorithm is about 3dB higher than that of the traditional soft threshold wavelet denoising. The root mean square error (RSME) is also reduced by about 30%，and the running time is reduced by 46% or so. Experimental results show that the algorithm is simple and fast in operation, and has important practical significance for the real-time noise elimination system of atmospheric environment monitoring.