High-resolution time of arrival estimation method for synthetic band chirp-signals using compressed sensing

A microwave radar system is a useful tool for terrain surface measurement even in adverse weather or darkness. Most important specification of a radar system is range resolution, which is strictly determined by the frequency bandwidth of transmitted signal. However, due to the legal regulation of radio wave emission, its frequency band is strictly limited. In the case that the maximum and minimum frequencies of the signal are fixed, the normal range resolution is same, but there is a severe problem of TOA ambiguity, when the effective bandwidth is considerably lower. To solve this problem, we introduce the compressed sensing (CS) approach. The results from numerical simulation show that our method can suppress the TOA ambiguity even that actual bandwidth of transmitted signal is quite narrower.

I. Introduction

Since a microwave radar system is one of the most useful tools for terrain surface observations being applicable to all-weather situations. It is promising for monitoring landslides after cataracs of rain or volcanic activity, where the optical sensor is hardly applicable [1]. The range resolution is one of the most significant specifications, which directly affects the spatial resolution of radar images. Usually, the range estimation issue is rephrased as the TOA (Time of Arrival) estimation one. For this background, a number of high-resolution TOA estimation approaches have been proposed, such as cross-correlation method, Capon method and MUSIC (MUltiple SIgnal Classification) method. Although the cross-correlation method is classical TOA estimation scheme, the range resolution is strictly limited by the transmitted frequency bandwidth [3]. The Capon method is one of the super resolution approaches, which can suppress the sidelobe response by minimizing the output power under the norm constraint for desired TOA [4]. While the MUSIC method is also a well known super resolution method, which is based on the eigenvalue decomposition of cross-correlation matrix, it needs a priori knowledge for the number of targets [5].