To solve the mismatching problem between the base dictionary and the signal in the existing multi-frequency hopping signal parameter estimation method, a time-frequency parameter estimation method based on atomic norm is proposed. First, the infinite atom set is defined and the atomic norm minimization model is established. Then, the dual problem is used to complete the signal estimation. Finally, the signal frequency is estimated based on the absolute value of the dual polynomial on the unit circle meanwhile transition moment of the signal is estimated by calculating the number of the frequency components. Simulation results show that this algorithm can obtain higher estimation accuracy and performance than existing algorithms under multi-frequency hopping signals.