1. INTRODUCTION
The task of blind source separation (BSS) is to recover N latent sources based on only observed data from M sensors, without knowing the mixing process and source information. Nowadays, BSS remains an active research topic with applications in a wide variety of fields, e.g., wireless communications [1], [2], biomedical [3], and image processing [4], [5]. The main difficulty in blind separation is the mixing matrix estimation (MME), which is of great importance to enable accurate reconstruction of original sources [6]-[18]. Although relevant methods and analysis assume a general setting, i.e., M ⩾ 2, this paper emphasizes on the particular case of a two-sensor array, i.e., M = 2.