This article is about the study of portable fiber optic spectrometer CO detection technology. The detection system were built by used small optical fiber spectrometer as the core ，and the spectrum signal used wavelet transform to preprocess. Processed CO spectral feature extraction based on the information using a gaussian fitting method and CO in the characterization of eigenvalues are obtained by fitting the spectral information， finally a precise mathematical model were established by using partial least squares method and principal component analysis (PCA) . The method can fast and extractly characterize information from spectral signal and build the CO detection model .