Although derivative can correct drift of spectra, it also brings on noise. The application of wavelet denoising (WD) to near infrared derivative spectra was investigated. The parameters such as wavelet function, threshold calculation and scale level were studied in detail. The WD performance was evaluated by means of ratio of signal-noise (S/N) and the predictive ability for RON (Research Octane Number) of gasoline. The results show that wavelet function and scale level have great effects on WD performance. WD can reduce markedly the noise from near infrared derivative spectra; improve effectively S/N and RON analysis accuracy. WD methods were compared with Fourier Transform denoising (FTD) and S-G smoothing (SGS) respectively. Wavelet methods are better than others are.
展开▼
