Acta Optica Sinica, Volume. 34, Issue 6, 630001(2014)
Peak Detection Algorithm of Raman Spectra Based on Multi-Scale Local Signal-to-Noise Ratio
Raman spectral peak recognition is one of the key technologies in qualitative analysis of Raman spectra. Due to the defects of low degree of automation and low recognition accuracy of the existing Raman spectral recognition methods, a new Raman peak recognition algorithm based on multi-scale local signal-to-noise ratio (MLSNR) is proposed. The algorithm gets the multi-scale second order difference coefficient of spectrum through multi-scale second order difference operation, then divides the multi-scale second order difference coefficient by the estimated noise standard deviation to obtain the MLSNR matrix of spectrum, and identifies Raman peaks by searching the ridges caused by local maxima in MLSNR matrix. The algorithm uses an automatic threshold estimation method to avoid the interference of local maximum caused by noise, and can recognize Raman peaks automatically without any parameter to be specified by human. The simulation result shows that no matter to singular peak or congested peaks, when the signal-to-noise ratio of Raman peak is greater than or equal to 6, the recognition accuracy of MLSNR algorithm is 100%, even to the singular peak at the detection limit, the recognition accuracy is more than 95%. MLSNR algorithm is a practical Raman spectral peak identification method.
Get Citation
Copy Citation Text
Jiang Chengzhi, Sun Qiang, Liu Ying, Liang Jingqiu, Liu Bing. Peak Detection Algorithm of Raman Spectra Based on Multi-Scale Local Signal-to-Noise Ratio[J]. Acta Optica Sinica, 2014, 34(6): 630001
Category: Spectroscopy
Received: Dec. 3, 2013
Accepted: --
Published Online: Apr. 23, 2014
The Author Email: Chengzhi Jiang (wangjcz@163.com)