1: \begin{abstract}

2: The least mean squares (LMS) filter is often derived via the Wiener filter solution. For a system identification scenario, 

3: such a derivation makes it hard to incorporate prior information on the system's impulse response. 

4: We present an alternative way based on the maximum a posteriori 

5: solution, which allows developing a Knowledge-Aided Kaczmarz algorithm. Based on this 

6: Knowledge-Aided Kaczmarz we formulate 

7: a Knowledge-Aided LMS filter. Both algorithms allow 

8: incorporating the prior mean and covariance matrix on the parameter to be estimated. 

9: The algorithms use this prior information in addition to the measurement information in the gradient for the 

10: iterative update of their estimates. We analyze the convergence of the algorithms and show simulation 

11: results on their performance. 

12: As expected, reliable prior information allows improving the performance of the algorithms for low signal-to-noise 

13: (SNR) scenarios. The results show that the presented algorithms can nearly achieve the optimal 

14: maximum a posteriori (MAP) performance.

15: \end{abstract}