1: \begin{abstract}

2: This paper investigates a channel estimator based on \acp{gmm} in the context of linear inverse problems with additive Gaussian noise.

3: We fit a \ac{gmm} to given channel samples to obtain an analytic \ac{pdf} which approximates the true channel \ac{pdf}.

4: Then, a \ac{cme} corresponding to this approximating \ac{pdf} is computed in closed form and used as an approximation of the optimal \ac{cme} based on the true channel \ac{pdf}.

5: This optimal \ac{cme} cannot be calculated analytically because the true channel \ac{pdf} is generally unknown.

6: We present mild conditions which allow us to prove the convergence of the \ac{gmm}-based \ac{cme} to the optimal \ac{cme} as the number of \ac{gmm} components is increased.

7: Additionally, we investigate the estimator's computational complexity and present simplifications based on common model-based insights.

8: Further, we study the estimator's behavior in numerical experiments including \ac{mimo} and wideband systems.

9: \end{abstract}

10: