A complex-valued signal is improper if it is correlated with its complexconjugate. The dimension of the improper signal subspace, i.e., the number ofimproper components in a complex-valued measurement, is an important parameterand is unknown in most applications. In this letter, we introduce twoapproaches to estimate this dimension, one based on an information- theoreticcriterion and one based on hypothesis testing. We also present reduced-rankversions of these approaches that work for scenarios where the number ofobservations is comparable to or even smaller than the dimension of the data.Unlike other techniques for determining model orders, our techniques also workin the presence of additive colored noise.
展开▼