A Common/Non-Common Space Separation Algorithm for Channel Characterization in Time-Varying Environments

With the rapid development of emerging wireless communication technologies, there is a concurrent surge in data volume alongside urging demands for real-time data processing in time-varying channels. Existing environment sensing methods can hardly realize real-time dynamic sensing due to the large amount of post-processing requirements on the static channel components. Therefore, the decomposition of static and dynamic channels has become a great challenge, contributing to efficient dynamic sensing. In this paper, a subspace-based method is proposed to separate the common and non-common components of the channel matrix, which is further used for static and dynamic channel sensing, respectively. The spatial and frequency subspaces of two channel state informations (CSIs) are extracted within the coherence time by eigenbasis projection, respectively, from which the common and non-common subspaces are separated to perform compressed sensing (CS) based CSI feedback. Spectrum-based analysis of the separated results by simulation and measurement is elaborated and the performance is numerically evaluated by correlation coefficients.