A compressed channel sensing method was proposed for Orthogonal Frequency Division Multiplexing (OFDM) based Amplify-and-Forward (AF) cooperative communication network over frequency-selective fading channels. First, by using cyclic matrix theory, the system model was established similar to the traditional point-to-point system model, which consisted of a cascaded channel vector and a measurement matrix. And then, using the theory of compressed sensing, the measurement matrix was proven to satisfy Restricted Isometry Property (RIP) with high probability. Finally, convolution channel impulse response was reconstructed with compressed sensing algorithm. According to the figures example, the cooperative channel exhibited an inherent sparse or sparse clustering structure. Hence, the proposed method can fully exploit the inherent sparse structure in cooperative channel. The simulation results confirm that the proposed method provides significant improvement in Mean Square Error (MSE) performance or spectral efficiency compared with the traditional linear channel estimation methods.