Robust phase shift estimation in noise for microphone arrays with virtual sensors

Virtualization is a new method which allows to build highly directive microphone arrays even if a small number of sensors is available. Using this method in the case of electronic hearing aids, non-physical (virtual) sensors can be created by having only a linear end-fire array of two physical microphones. Then the global array processing with more than two sensors can be implemented in the time or frequency domain. In this paper, frequency domain techniques, which include the application of the discrete Fourier transform (DFT) and phase shift beamforming are considered. Our prime interest is in the phase shift estimation for virtual microphones in the presence of acoustic background noise. A careful consideration of the non-stationary nature of the signals and the noise influence allows the design of robust frequency domain algorithms to be easily integrated into a complete and efficient hearing aid system.