Most conventional methods and algorithms for direction finding suffer from poor accuracy when subjected to scenarios characterized by closely spaced sources, short data records and correlated source signals. In the last few years, some approaches to overcoming these problems have exploited prior knowledge of signal directions coming from static users. However, this concept is restricted to known directions of arrival. This thesis presents several contributions to dealing with the aforementioned problems. A novel concept of a priori knowledge applied to direction finding is first presented, which replaces the traditional available known DOAs so far employed with previous estimates obtained on line. This idea is then incorporated into the proposed algorithms and their extensions. Another approach is also introduced to estimating the data covariance matrix by iteratively reducing its by-products, which occurs in the finite sample region. This concept is complemented by a reshaped covariance matrix analysis, which shows that after the first iteration the Mean Squared Error of the data covariance matrix free of these side effects is less than or equal to the MSE of the original one. A novel method, termed Multi-Step Knowledge-Aided Iterative (MS-KAI), for increasing the accuracy of existing algorithms based on the combination of the previous concepts is then developed. The MS-KAI method is initially employed with Uniform Linear Arrays (ULAs) and is combined with the Estimation of Signal Parameters via Rotational Invariance Techniques algorithm, resulting in the proposed MS-KAI-ESPRIT algorithm. Then, MS-KAI is extended for use with an arbitrary number of iterations and combined with the Conjugate Gradient algorithm, resulting in the MSKAI- CG algorithm. Finally, the MS-KAI method is considered with nested arrays and combined with the Multiple Signal Classification algorithm, resulting in the proposed MS-KAI-MUSIC algorithm. Simulation results show that MS-KAI method enhances the accuracy of subspace based algorithms employing ULA and non-ULA based system models.