Blind adaptive filtering of speech from noise of unknown spectrum using a virtual feedback configuration

The paper describes a single-receiver blind adaptive filter (BAF) of speech from noise where neither speech nor noise are accessible, nor are their parameters known. The only prior knowledge employed by the BAF is that human speech is nonstationary whereas the noise is assumed to be quasistationary, i.e., stationary over a longer interval than that of any speech phoneme. The RAF has a four subsystem structure. The system consists of an identifying subsystem that is followed by a speech/noise parameter-separator. The noise is identified based on the stationary features of speech and noise. A feedforward subsystem sets optimization neighborhood to the virtual feedback subsystem where a cost-functional is minimized to jointly minimize the stationary part of the output while maximizing its nonstationary part. The system has been tested for performance for different signal to noise ratios (SNR) and for different types of noise parameters. Improvements for various noises range from 14-36 dB for -20 dB SNR inputs.