ABSTRACT
This chapter assumes that an estimate of the noise spectrum is available. Such an estimate is critical for the performance of speech-enhancement algorithms as it is needed, for instance, to evaluate the Wiener filter in the Wiener algorithms or to estimate the a priori signal-to-noise ratio in the minimum mean square error algorithms or to estimate the noise covariance matrix in the subspace algorithms. The simplest approach is to estimate and update the noise spectrum during the silent segments of the signal using a voice activity detection (VAD) algorithm. The chapter focuses on a new class of algorithms that estimate the noise continuously, even during speech activity. The majority of the VAD algorithms encounter problems in low-signal-to-noise ratio conditions, particularly when the noise is nonstationary. Noise-estimation algorithms that continuously track the noise spectrum are therefore more suited for speech-enhancement applications in nonstationary scenarios.
