REAL-TIME TF-GSC IN NONSTATIONARY NOISE ENVIRONMENTS

Adaptive beamforming techniques are inefficient for eliminat-ing transient noise components that randomly arrive from unpre-dictable directions. In this paper, we present a real-time transfer function generalized sidelobe canceller (TF-GSC) for such non-stationary noise environments. Hypothesis testing in the spectral domain indicates either absence of transients, presence of an in-terfering transient, or presence of desired source components. The noise canceller branch of the TF-GSC is updated only during ab-sence of transients, while the identification of the acoustical trans-fer function is carried out only when desired source components are present. Following the beamforming and the hypothesis test-ing, estimates for the signal presence probability, the noise power spectral density, and the desired speech log-spectral amplitude are derived. Experimental results demonstrate the usefulness of the proposed approach under nonstationary noise conditions. 1