Abstract: ‍ ‍Aimed at the problem of insufficient bandwidth and power tolerance of a single cover signal in the traditional satellite overlapping communication， a multi-cover signal overlapping communication method with frequency domain division and sub-spectrum power control joint optimization was proposed by using multiple cover signals in the satellite transponder spectrum environment. A bi-objective optimization problem of signal transmission performance and stealth performance for covert communication was established. The gateway side used the perceived historical spectrum data to train and generate the support vector machine regression prediction model， which was used to predict the communication performance and stealth performance of covert signals in different transponder spectrum environments， and downloaded the trained prediction model to the communication terminal. The terminal side used the dual-target knapsack algorithm to select cover signals in the transponder spectrum environment by taking the communication performance and covert performance of the covert signal predicted by the support vector machine regression prediction model as value factors， and the number of cover signals as the backpack weight， and solved the frequency domain division of the concealed signal and the power control parameters of the sub-spectrum， so as to realize the purpose of hiding the terminal communication signal in the frequency spectrum environment of the satellite transponder. This method not only solves the problem of insufficient power and bandwidth tolerance of single cover signal in traditional covert communication， but also comprehensively optimizes the communication performance and covert performance of covert communication system. Due to the long propagation delay of satellite communication， the advantage of adopting the method proposed in this paper is that the terminal and the gateway station can avoid the interaction between the receiver and the receiver in the communication process by using the same historical database and the machine learning prediction model， and quickly obtain the required covert communication transmission strategy based on the current spectrum environment.