| The important trend of spectrometer is microminiaturization,integration and intelligence. The development of micro optics electronics mechanics system(MOEMS) creates condition for microminiaturization and integration of spectrometer, the development of computer,artificial intelligence,chemometrics offers opportunity for intelligence of spectrometer.Spectral signal analysis is not only an important part of the analysis system of spectrometer, but also key to intelligence of spectrometer. The research on spectral signal analysis is useful to advance performance of spectrometer and extend application range of spectrometer.With the support of the Chinese National Nature Science Fund,863 Fund,the Department of Science and Technology Fund, the paper takes spectral signal as object, studies the technology of spectral signal analysis about the micro spectrometer of the MEMS center of Chong qing University. The primary coverage is listed as following.①In this paper, a frame of functional model about software system is built. Every functional model is designed in detail.②The method based on uniformity design,wavelet analysis,cross validation and support vector machine is put forward for solve the question of nonlinearity,small-sample in the spectral multicomponent analysis.③The method of selecting good wavelength data for building analysis model is an effective improving analysis precision.To quicken convergence, the piecewise wavelength selection method based on particle swarm optimization algorithm is established.④The technology of spectral recognition is key to spectral quapitative analysis.For quickly recognising spectral signal,the method of spectral recognition based on based on multiple features and neural network is adopted.⑤The model transfer is a basic method to solve universal and comparable performance of spectrometers.The method based on support vector machine and piecewise direct standardization is put forward for solve the question of nonlinearity,small-sample.
|
PDF Full Text Request