Blind Source Separation of Single-Channel Background Sound Cockpit Voice Based on EEMD and FastICA

In order to solve the problem of extremely underdetermined blind source separation (BSS) with only one dimensional observing matrix, an ensemble empirical mode decomposition (EEMD) method is proposed to decompose the single-channel mixed signal into multiple instantaneous frequency intrinsic mode function (IMF). The new observing matrix is constructed, and then the blind source separation is realized by fast independent component analysis (FastICA). Simulation experiments and laboratory studies show that this method can suppress broadband and transient interference, and effectively extract the target signal submerged in noise. The analysis of the measured data shows that the method can effectively extract the background sound cockpit voice signal under the interference of aircraft engine noise, which proves the effectiveness of the method in the cockpit voice signal processing.