Corcoran Timothy C
306571 Department of Chemistry and Biochemistry, California State Polytechnic University Pomona, Pomona, CA, USA.
Appl Spectrosc. 2018 Mar;72(3):392-403. doi: 10.1177/0003702817738023. Epub 2017 Nov 28.
In the chemometric context in which spectral loadings of the analytes are already known, spectral filter functions may be constructed which allow the scores of mixtures of analytes to be determined in on-the-fly fashion directly, by applying a compressive detection strategy. Rather than collecting the entire spectrum over the relevant region for the mixture, a filter function may be applied within the spectrometer itself so that only the scores are recorded. Consequently, compressive detection shrinks data sets tremendously. The Walsh functions, the binary basis used in Walsh-Hadamard transform spectroscopy, form a complete orthonormal set well suited to compressive detection. A method for constructing filter functions using binary fourfold linear combinations of Walsh functions is detailed using mathematics borrowed from genetic algorithm work, as a means of optimizing said functions for a specific set of analytes. These filter functions can be constructed to automatically strip the baseline from analysis. Monte Carlo simulations were performed with a mixture of four highly overlapped Raman loadings and with ten excitation-emission matrix loadings; both sets showed a very high degree of spectral overlap. Reasonable estimates of the true scores were obtained in both simulations using noisy data sets, proving the linearity of the method.
在化学计量学背景下，若分析物的光谱载荷已知，则可构建光谱滤波函数，通过应用压缩检测策略，直接以即时方式确定分析物混合物的得分。无需在相关区域收集混合物的完整光谱，而是可在光谱仪自身内部应用滤波函数，从而仅记录得分。因此，压缩检测极大地缩减了数据集。沃尔什函数是沃尔什 - 哈达玛变换光谱学中使用的二进制基，构成了非常适合压缩检测的完备正交归一集合。借鉴遗传算法的数学方法，详细阐述了一种使用沃尔什函数的二进制四重线性组合构建滤波函数的方法，以此针对特定分析物集优化所述函数。这些滤波函数可构建为能自动从分析中去除基线。针对四种高度重叠的拉曼载荷混合物以及十种激发 - 发射矩阵载荷进行了蒙特卡罗模拟；两组数据均显示出非常高的光谱重叠度。在两个模拟中使用有噪声数据集均获得了对真实得分的合理估计，证明了该方法的线性。
