Abstract: A satellite spectrum multi-threshold sensing algorithm based on long short-term memory (LSTM) neural network is proposed in allusion to the problems of low sensing performance and high influence of communication delay in satellite cognitive communication.Firstly，the satellite cognitive communication model is constructed，then the simulation data is sent to the LSTM neural network for prediction and sensing，and the network is updated by the stochastic gradient descent with momentum (SGDM) algorithm. Then the multi-threshold algorithm is proposed to optimize the output of LSTM neural network，and finally，the performance is compared with other neural network algorithms.The results show that under the condition of satellite channel，the spectrum sensing performance of the proposed algorithm is better than that of other neural network algorithms in low signal-to-noise ratio (SNR) and low neural network iterations.