This paper presents a new frequency domain blind signal extraction (FD-BSE) method for the extraction of a target speech in presence of diffuse background noise. This is a fast alternative to frequency domain blind signal separation (FD-BSS) for hands-free speech interface. Like the FD-BSS approach, the speech signal is enhanced by using a nonlinear filter to suppress the noise estimated by the blind method. Simulation results in a realistic environment show the effectiveness of the proposed method.