Deep Learning Detection for superimposed control signal in LEO-MIMO

Multiple-input multiple-output (MIMO) communication systems using multiple low-Earth orbital (LEO) satellites achieve higher capacity than conventional LEO systems. However, in previous research, control signals are allocated to a different frequency band for each satellite signal in order to estimate its Doppler frequency. The increase in the number of satellites reduces unoccupied MIMO signal bandwidth and hence the overall capacity. This study aimed to prevent such capacity reduction by introducing the superimposition of control signals. Such control signals occupy a frequency bandwidth of the equivalent of only one control signal; therefore, they can prevent the reduction of the overall capacity. The main challenge is estimating the satellites' Doppler frequencies from the waveforms of Doppler-affected superimposed control signals. To overcome this challenge, we propose the adoption of a deep learning technique.