Skip to main content

Table 3 Conditions for SS-based dereverberation

From: Dereverberation and denoising based on generalized spectral subtraction by multi-channel LMS algorithm using a small-scale microphone array

Analysis window Hamming
Window length 32 ms
Window shift 16 ms
Number of reverberant windows D 6
  (192 ms)
Noise overestimation factor α 1.0 (Power SS)
  0.1 (GSS)
Spectral floor parameter β 0.15 (both)
Soft-mask gradient parameter a 0.05 (Power SS)
  0.01 (GSS)
Soft-mask center parameter b 0.0 (both)