Aiming at the characteristics of nonlinearity，non-stationary and transient impact of telemetry vibration signal，a telemetry vibration signal feature enhancement method based on time-frequency manifold adaptive sparse reconstruction was proposed.The phase space reconstruction of the vibration signal was performed to extract its time-frequency manifold. Based on the time-frequency manifold，the KSVD algorithm was used to adaptively construct an over-complete dictionary and find the best matching time-frequency atoms. According to the selected atoms，the time-frequency distribution of each dimension signal in phase space was sequentially matched to calculate the reconstructed sparse coefficient. The time-frequency distribution of each dimension signal in phase space was reconstructed by using sparse coefficient and time-frequency atoms，and the feature enhancement signal was acquired by inverse operation of time-frequency distribution and phase space reduction. Simulation and measured signal processing results verified the effectiveness of the method.