This paper describes an algorithm for robust time-delay estimation (TDE)
in situations where a large amount of additive noise and reverberation
is present. In  an adaptive eigenvalue decomposition algorithm has
been developed for TDE between two microphones in highly reverberant
acoustic environments. In this paper we extend this algorithm to highly
noisy and reverberant environments, by using a generalized eigenvalue
decomposition or by prewhitening the noisy microphone signals. It is
shown that time-delays can be robustly estimated for SNRs down to -5 dB.