光譜的量測很容易受到雜訊影響，所以當訊號有微小的變化時，可能會被誤判為雜訊，而造成訊號的失真，所以雜訊濾除是一項非常重要的工作，一開始將輸入的光譜訊號分類，分成線光譜、寬帶光譜、線光譜混合寬帶光譜，線光譜使用Gaussian fitting，寬帶光譜使用S-G濾波器進行濾波，而S-G濾波器參數使用FOA演算法自動尋找最佳參數，再將每次找到的參數畫做圖，並找出一個代表點，以後進行S-G濾波時即可迅速完成，遇到線光譜混合寬帶光譜時則使用分開做濾波處理，最後再合併的方法進行濾波達成目的，由實驗模擬結果可知，利用RMS尋找代表點可得到光譜儀取樣間距為0.2 nm，多項式階數8，取樣個數129為最佳，光譜儀取樣間距為1 nm，多項式階數8，取樣個數39為最佳，光譜儀取樣間距為3 nm，多項式階數9，取樣個數19為最佳，所以只要搭配已知硬體的取樣間距，即可快速完成S-G濾波。

Measurement of the spectrum is susceptible to noise impact easily, therefore, when the signal slightly changes, it will be misunderstand as noise, causing signal distortion, thus, the noise filter is an important job, initial input spectral signal can be classified into line spectrum, broad-band spectrum, mixed the broad-band spectrum and line spectrum, line spectrum uses the Gaussian fitting, the broad-band spectrum uses S-G filter for filtering, the S-G filter parameters uses the FOA algorithm to find the optimal parameters automatically, then find each parameter to draw a map and find a representative point, hence S-G filter can be completed quickly, mixed the broad-band spectrum and line spectrum should separate to deal with the filter, finally, using merged method to proceed filter and achieve this objective, the experiment simulated result indicates that using RMS to find representative points can know spectrometer resolution is 0.2 nm, polynomial order is 8，moving window of 129 is the best, spectrometer resolution is 1 nm, the polynomial order is 8, moving window of 39 is the best, spectrometer resolution is 3 nm, polynomial order is 9, moving window of 19 is the best, so with the known resolution of hardware, S-G filter can be quickly completed.

[1]John W. Jewett, Jr. and Raymond A. Serway, Physics for Scientists and Engineers with Modern Physics, 8th ed. Canada: Mary Finch, 2010.
[2]林耀禎，「以Savitzky Golay演算法分析SNR與雜訊容忍度制定訊號敏感度區段方法研究」，碩士論文，自動化及控制研究所，國立台灣科技大學，台北，2011。
[3]Sir Isaac Newton， http://zh.wikipedia.org/wiki/%E8%89%BE%E8%90%A8%E5%85%8B%C2%B7%E7%89%9B%E9%A1%BF。
[4]Wen-Tsao Pan, “A New Fruit Fly Optimization Algorithm:Taking the Financial Distress Model as an Example,” Knowledge-Based Systems, vol. 26, pp. 69-74, 2011.
[5]江淑敏，「修飾雜訊光譜配合匹配濾波器與二次微分濾波器改善液相層析質譜訊號峰之訊雜比」，碩士論文，化學所，中正大學，嘉義，2006。
[6]國立高雄師範大學-科學教育中心，http://140.127.36.251/sc_scedu_knowledge.htm。
[7]魏克新，MATLAB語言與自動控制系統設計，北京，機械工業出版社，1997.8。
[8]洪維恩，Matlab7程式設計，旗標出版股份有限公司，2005.6.21。
[9]廣義平均，http://en.wikipedia.org/wiki/Generalized_mean#Definition。
[10]Savitzky, A.; Golay, M. J. E., “Smoothing and Differentiation of Data by Simplified Least Squares Procedures,” Analytical Chemistry, 36 (8), pp. 1627–1639, 1964.
[11]潘文超，果蠅最佳化演算法─最新演化式計算技術，滄海書局，鼎隆圖書股份有限公司，2011。