Ideal Binary Mask

Applies ideal binary mask to noise corrupted speech signal within an STFT framework.

Applies ideal binary mask to noise corrupted speech signal. The processing takes place within an FFT based short-time spectral analysis-modification-synthesis framework. The ideal binary mask is computed from an oracle (true) signal-to-noise ratio (SNR) through thresholding with local SNR criterion. The routine returns synthesized enhanced speech, the ideal binary mask and the true instantaneous spectral SNR.