High-resolution Direction-of-Arrival estimation

Direction of Arrival (DOA) estimation is considered one of the most crucial problems in array signal processing, with considerable research efforts for developing efficient and effective direction-finding algorithms, especially in the transportation industry, where the demand for an effective, real-time, and accurate DOA algorithm is increasing. However, challenges must be addressed before real-world deployment can be realised. Firstly, there is the requirement for fast computational time for real-time detection. Secondly, there is a demand for high-resolution and accurate DOA estimation. In this thesis, two state-of-the-art DOA estimation algorithms are proposed and evaluated to address the challenges. Firstly, a novel covariance matrix reconstruction approach for single snapshot DOA estimation (CbSS) was proposed. CbSS was developed by exploiting the relationship between the theoretical and sample covariance matrices to reduce estimation error for a single snapshot scenario. CbSS can resolve accurate DOAs without requiring lengthy peak searching computational time by computationally changing the received sample covariance matrix. Simulation results have verified that the CbSS technique yields the highest DOA estimation accuracy by up to 25.5% compared to existing methods such as root-MUSIC and the Partial Relaxation approach. Furthermore, CbSS presents negligible bias when compared to the existing techniques in a wide range of scenarios, such as in multiple uncorrelated and coherent signal source environments. Secondly, an adaptive diagonal-loading technique was proposed to improve DOA estimation accuracy without requiring a high computational load by integrating a modified novel and adaptive diagonal-loading method (DLT-DOA) to further improve estimation accuracy. An in-depth simulation performance analysis was conducted to address the challenges, with a comparison against existing state-of-the-art DOA estimation techniques such as EPUMA and MODEX. Simulation results verify that the DLT-DOA technique performs up to 8.5% higher DOA estimation performance in terms of estimation accuracy compared to existing methods with significantly lower computational time. On this basis, the two novel DOA estimation techniques are recommended for usage in real-world scenarios where fast computational time and high estimation accuracy are expected. Further research is needed to identify other factors that could further optimize the algorithms to meet different demands