Estimate the degree of noncircularity of complex-valued vectors via exponentially embedded family

Document Type

Conference Proceeding

Date of Original Version



Complex-valued signal processing is a fundamental task in many signal processing areas such as radar, sonar and communications. Modeling complex data as noncircular may provide better fitting of physical conditions. However, it requires more complicated signal processing algorithms and hence has more computations. Testing of noncircularity and estimating its degree are helpful in choosing a model. In this paper we focus on estimating the degree of noncircularity if the data is decided noncircular. It essentially a model order selection problem; therefore, we adopt the recently proposed exponentially embedded family (EEF) rule. Computer simulations are given to evaluate the EEF's performance and compare it with the minimum description length (MDL).

Publication Title

2017 IEEE Radar Conference, RadarConf 2017