Method and system for extracting material parameters in terahertz time-domain spectrometer

The invention discloses a method and a system for extracting material parameters in a terahertz time-domain spectrometer. The method comprises the following steps: acquiring a terahertz time-domain waveform of a to-be-detected material; according to a preset material parameter extraction network, performing parameter extraction on the terahertz time domain waveform to obtain physical parameters of the to-be-detected material; wherein the preset material parameter extraction network is obtained by transmitting a reference signal and a sample signal corresponding to different material parameters simulated by a physical model into a convolutional neural network for training; the sample signal is obtained by processing a reference signal and a material characteristic matrix through a migration matrix method; the reference signal is obtained through bilateral exponential function modeling. According to the invention, the feature representation of the terahertz signal is automatically learned through the preset mat