Many classical chirp signal processing algorithm may experience distinct performance decrease in noise circumstance. To address the problem, this paper proposes a deep learning based approach to filter noises in time domain. The proposed denoising convolutional neural network (DCNN) is trained to recover the original clean chirps from observation signals with noises. Following denosing, we employ two parameter estimation algorithm to DCNN output. Simulation result show that the proposed DCNN method improves the signal noise ratio (SNR) and parameter estimation accuracy to a great extent compared to the signals without denoising. And DCNN have a strong adaptability of low SNR input scenarios that never trained.