Deep learning is a neural-inspired pattern recognition technique that has been shown to be as effective as conventional signal processing and has considerable potential to identify gravitational-wave signals. In this talk, I will first review some works on the detection and characterization of gravitational-wave signals, then present our related progress about the effect of deep neural networks on the recognition and identifying generalization properties of gravitational waves. After that, based on a proposed idea from matched-filtering method, we design a brand new architecture (MF-CNN) and . At last, I will explain some details of our experiments and the main properties on the real LIGO recordings.