机构地区: 广东科学技术职业学院
出 处: 《光谱实验室》 2005年第3期656-659,共4页
摘 要: 以测定的紫丁香醛近红外漫反射光谱信号为标准,随机加入了一噪音信号,采用Daubechies小波系中的DB5小波函数,以4种阈值选取方法(最小最大准则阈值、Stein无偏似然估计阈值、混合阈值和固定阈值)和2种阈值函数(软阈值函数和硬阈值函数)组合对含噪音的光谱信号进行了小波去噪处理试验,以均差MSE评价组合各种阈值选取及阈值函数方法的去噪优劣。结果表明,对于该类型的光谱信号,采用最小最大阈值和软阈值函数组合,得到最优的去噪效果。 A random noise was added to a standard near-infrared diffuse reflectance spectrum of syringaldehyde, and the wavelet mother db5 was used to de-noise the spectrum containing the noise with the four threshold selection rules(mini-maxi,rigrsure,heursure and sqtwolog) ,and two threshold functions (soft and hard). Mean squared error (MSE) was used to estimate these performances. The best de-noising performance is the combination of 'mini-maxi' threshold and soft threshold function.
领 域: [理学] [理学] [自动化与计算机技术] [自动化与计算机技术]