On Improvements of Cyclic MUSIC

  • PDF / 759,293 Bytes
  • 8 Pages / 600 x 792 pts Page_size
  • 66 Downloads / 215 Views

DOWNLOAD

REPORT


On Improvements of Cyclic MUSIC Huiqin Yan Department of Electrical and Computer Engineering and Computer Science (ECECS), University of Cincinnati, Cincinnati, OH 45221-0030, USA Email: [email protected]

H. Howard Fan Department of Electrical and Computer Engineering and Computer Science (ECECS), University of Cincinnati, Cincinnati, OH 45221-0030, USA Email: [email protected] Received 20 December 2003; Revised 10 June 2004 Many man-made signals encountered in communications exhibit cyclostationarity. By exploiting cyclostationarity, cyclic MUSIC has been shown to be able to separate signals with different cycle frequencies, thus, to be able to perform signal selective direction of-arrival (DOA) estimation. However, as will be shown in this paper, the DOA estimation of cyclic MUSIC is actually biased. We show in this paper that by properly choosing the frequency for evaluating the steering vector, the bias of DOA estimation can be substantially reduced and the performance can be improved. Furthermore, we propose another algorithm exploiting cyclic conjugate correlation to further improve the performance of DOA estimation. Simulation results show the effectiveness of both of our methods. Keywords and phrases: direction-of-arrival estimation, uniform linear array, cyclic MUSIC, cyclostationarity.

1. INTRODUCTION Among subspace-based DOA estimation methods, MUSIC [1, 2] is relatively simple and effective, and is thus the most studied. To improve the performance of the conventional MUSIC, cyclic MUSIC [3] is shown to be effective to combat noise and interference by exploiting cyclostationary property possessed by most man-made signals in communications [4]. From then on, many papers have been developed to improve the performance of cyclic MUSIC, such as [5, 6, 7]. Reference [5] is asymptotically exact for either narrowband or wideband sources, but as stated in [5], as long as the crosscyclic correlations of the sources are not small enough, they will act consistently as small induced interferences with the same cycle frequency, especially when the source is narrowband. Thus, this method works better for the wideband case. By narrowband, we mean that the signal “fractional bandwidth,” that is, the ratio of the signal bandwidth over its carrier frequency, is small. Reference [6] exploits both cyclostationarity and conjugate cyclostationarity to improve the ability of separating two closely impinging signals, and in certain condition, it can detect more signals than the number of antennas. Reference [7] makes use of spatial smoothing (SS) [8], and presents a scheme called Hankel approximation method (HAM). Then in conjunction with cyclic MUSIC, it solves the DOA estimation problem in the presence of coherent signals. Both [6, 7] are based on the assumption that the

signals are narrowband and the model x(t) = As(t) + n(t) holds exactly. Many papers refer to this model as a narrowband model; whereas, it actually holds exactly only for pure sinusoidal signals (see more details in Section 3). Thus in this paper, we will refer to thi