PROCESSING OF NOISY SPEECH USING PARTIAL PHASE

This paper explores the possibility of processing noisy speech using signal reconstruction algorithms from Fourier Transform (FT) phase and magnitude. Algorithms have been proposed in the literature for signal reconstruction from FT phase alone/or, from FT magnitude with additional information in the form of 1-bit phase or signal values. More recently/algorithms have been proposed for signal reconstruction from partial phase (phase information in selected frequency bands) with compensating number of signal samples. In this paper we examine application of these techniques for processing noisy speech. In particular/we show that by selectively processing high signal-to-noise ratio(SNR) regions we can reduce the effect of background additive noise significantly.