Abstract: In order to solve the problem of traditional radar detecting performance has been greatly influenced by the relevant clutter, this paper proposes a sequential signal detection method for cognitive radar. The method divides the whole work time into two stages：Estimating and detecting. In the estimating stage, the optimal estimating waveform is launched and the target impulse response is estimating updated. In the detecting stage, system designs the optimal detecting waveform based on the prior knowledge and adaptively adjusts the detecting threshold. The results show that: The detecting performance of system has been improved obviously because of the joint optimization between the transmitter and receiver. In addition, the estimating accuracy of target impulse response increases when the number of loop increases so as to the estimated detecting probability will approach the true detecting probability, which would improve system performance and ensure the reliability of the decision at the same time.