Ambient backscatter is a new technology that utilizes environmental wireless signals to enable battery-free devices to communicate with each other. In this paper, we investigate the problem of signal detection for the full duplex wireless system utilizing ambient backscatter. Since one of the key technologies of full duplex wireless system is self-interference cancellation, we first estimate the channel between transmit antenna and receive antenna to eliminate self-interference. Next, a maximum likelihood (ML) detector is proposed to decode the signals from the ambient backscatter device, and the theoretical and practical thresholds are derived in the presence of channel estimation error. In addition, we demonstrate the existence of the optimal number of training symbols that can maximize the system throughput. Finally, simulation results are provided to corroborate our theoretical studies.