Study On Wavelet Analysis And Feature Extraction For Near-infrared Spectrum Modeling

With the complexity of the difficulties to resolve, complex samples by near infrared Spectrum is a high-dimensional complex data. It is the key of the near-infrared spectroscopy modeling in eliminating the impact of useless information from high-dimensional spectral data to extract key features effectively, which are the scientific problems and technical difficulties. Therefore, it is important to study the information extraction and processing technology of the near-infrared spectral features.The important feature of wavelet analysis is local time-frequency. Because the signal can be split according to different scales by this analysis, so they are widely used in all areas of information processing. It is an important means to achieving the feature extraction by processing or filtering in frequency-domain information, particularly the process of wavelet decomposition and reconstruction. This article is studying that how to improve the model accuracy by wavelet feature and modeling between the near-infrared spectroscopy and water for transmission. As the feature extraction is a hard NP problem usually in mathematics, so in this paper, we use the method of coarse-grained information. In a supervised learning mode, we extract the key features of information, optimize the model, improve the precision of the model by selecting the wavelet coefficients.