Skip to main content

Table 7 Breakdown of speech recognition errors (%)

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

  CMN only Power SS GSS (proposed)
   w/o MFT MFT w/o MFT MFT
Sub 40.61 30.48 29.37 27.39 27.42
Del 13.82 9.27 9.26 8.99 8.06
Ins 3.67 4.44 4.10 3.47 3.67