The respiratory measurement and classification system based on fiber Bragg grating is studied in this paper. In order to facilitate the needs of intelligent wear, the bare fiber grating was encapsulated with polydimethylsiloxane (PDMS), and the respiratory monitoring system was built to measure the respiratory signal. Four kinds of respiratory signals including breath-holding, cough, normal breathing and post-exercise breathing were collected. Based on wavelet decomposition and reconstruction, the collected respiratory signals were preprocessed and the frequency, amplitude factor, waveform factor and energy of the respiratory signals were extracted as characteristics to distinguish respiratory types. A respiration classification model based on support vector machine (SVM) was constructed, and the model parameters of SVM were optimized by particle swarm optimization. Finally, the classification accuracy was achieved at 97.1875%. The system is characterized by low cost, compact structure and simple design, which can enrich the digital diagnosis and treatment technology.