The Relative Transfer Matrix (ReTM) is a generalization of the Relative Transfer Function (ReTF) to scenarios involving three or more receivers and multiple simultaneously active sources. The ReTM allocates receivers into two multi-channel groups A and B, and formulates the spatial mapping between them for given sources. The conventional ReTM relies on a stationary environment to estimate the underlying covariance matrices. However, for dynamic environments where the microphone array is moving (e.g., head-mounted array) and/or the sources are moving, an adaptive estimation technique is needed. This paper introduces an Extended Kalman Filter (EKF) framework that formulates the ReTM estimation as a dynamic state-space model. We apply and evaluate the proposed approach to a scenario with a slowly moving head-mounted microphone array. Simulation results demonstrate that the EKF-based adaptive ReTM estimation is effective in dynamic environments and the EKF reliably tracks temporal variations.