Proceedings of Asia-Pacific Microwave Conference 2007 Unequal Spacing and Reference Element Variation To Enhance Resolution Of Linear Prediction DOA Algorithm Reza Abdolee, Mohd Nor Md Tan, Tharek Abd Rahman, Mohd Tarmizi Ali Wireless Communication Center (WCC), Universiti Teknologi Malaysia, Skudai Johor, Malaysia Email: reza.ab@ieee.org Abstract: In this paper, we proposed an unequal spacing technique which can increase the performance resolution of direction of arrival (DOA) estimation algorithms. The linear prediction (LP) algorithm has been tested for the proposed antenna arrangement, dramatic improvement has been obtained. In addition, we investigate the effect of reference element position on the performance resolution of the LP algorithm. It has been found that the performance resolution of the system increases if the reference element is chosen either at the first or last element of the linear arrays. As a conclusion of the work, the performance resolution of M elements antenna with unequal spacing is almost the same as 2M elements antenna with equal spacing. Key words: DOA estimation, unequal spacing, uniform linear array, Linear Prediction Algorithm 1. Introduction One of the important parts in the smart antenna system is the DOA estimation algorithm, the information from this part is fed to the adaptive algorithms to calculate the weights. The location of any user is continuously tracked by DOA algorithms. Various algorithms can be applied in this system in order to track the signal continuously such as Multiple Signal Classification (MUSIC) [1], Spectral Estimation Methods (SEM) [2], Linear Prediction Method (LP) [3], Estimation of Signal Parameters via Rotational Technique (ESPRIT) [4], Minimum Variance Distortionless Response (MVDR) [5], Maximum Entropy Method (MEM) [6] and Maximum Likelihood Method (MLM) [7]. Technically, the resolution of the DOA algorithms is the dominant factor to judge the performance of those algorithms. The researches trend in DOA algorithms is to increase the speed and the performance resolution of DOA estimation. The most popular DOA algorithms take advantages of correlation matrix and eigen decomposition of input data. However, the performance resolution of those algorithms is not so good and the performance degrades in noisy and correlated source signal. Subspace and eigen decomposition algorithms like MUSIC algorism also suffer from intensive computations. Although there are some methods to decrease the rate of computational complexity but yet they are not simple enough to be applied in the real applications. Therefore, making balance between two factors, the performance resolution and simplicity have a high importance in smart antenna systems. As it mentioned before, the performance resolution of all these algorithms will degrade in noisy environment. The performance also can be affected by the array correlation matrix. Therefore the higher degree of diversity is needed to improve the system performance. As a rule of thumb, when the number of antennas in the system increases, the performance resolution of the system increases. However, by doing that the system cost and complexity will increase significantly proportional with number of antenna elements. In this paper, we propose the unequal linear array structure which can dramatically increase the performance resolution of the system with lesser number of antenna. To do so, linear prediction methods as a representative of DOA estimation is chosen to verify this idea. Linear prediction method is one of DOA algorithm which its objective function is based on minimizing the predicted error and increase the signal to noise ratio of the system. The LP algorithm is applied to space domain DOA estimation by Burg [8] in 1967. Although LP algorithm not optimal technique but it is an attractive method because of its good prediction performance and relatively low computational complexity. 2. System model Consider a M-element un-equally spaced linear array, which is illustrated in Figure 1. The array elements are divided into 3 groups. Each group has M13 number of elements. The inter element spacing for each group are d, 2d and 4d for group-l,group-2 and group-3 respectively. The value of d is half of wavelength. 1-4244-0749-4/07/$20.00 @2007 IEEE.