Abstract: A matching pursuit method is used for Lamb wave signal processing so as to remove the redundancies of the detection signals and effectively identify the multiple Lamb wave modes.The Chirplet atoms are improved by adding a bending operator to match the dispersion and multiple mode characteristics of Lamb waves.Using the sparse decomposition of matching pursuit method based on genetic algorithms,the best matched atoms are selected from the over-complete dictionary which consists of improved Chirplet atoms.The Lamb waves are reconstructed and processed by time-frequency analysis from the best atoms and their corresponding decomposition coefficients.The study demonstrates that the matching pursuit method based on the improved Chirplet atoms can better reflect the nonlinear time-frequency change characteristics of Lamb wave detection signals,and the time-frequency distributions of Lamb wave signals agree with the bending characteristics of the dispersion curves very well.Thus,the matching pursuit method based on the improved Chirplet atoms could get the time-flight information of Lamb waves more accurately,which lays the foundation for further structural damage localization imaging using Lamb waves.