Skip to main content
EURASIP Journal on Audio, Speech, and Music Processing

Table 2 Worst-case scenario for a single CD codebook under babble noise

From: Robust Bayesian estimation for context-based speech enhancement

 

ΔPESQ

ΔSSNR (in dB)

Input SSNR

0 dB

5 dB

10 dB

0 dB

5 dB

10 dB

CBSE with CD

0.14±0.09

0.12±0.05

0.07±0.05

4.52±0.90

3.72±0.69

1.75±0.68

CBSE with CI

0.17±0.07

0.20±0.06

0.23±0.05

5.51±0.78

5.01±0.74

3.53±0.63

Proposed

0.17±0.07

0.20±0.06

0.21±0.06

5.52±0.79

4.98±0.73

3.47±0.62

Reference

0.09±0.02

0.12±0.03

0.15±0.05

2.40±0.40

2.99±0.40

3.09±0.44

  1. The CD codebook is modeling two aspects of context, speaker and acoustic environment. Both mean and standard deviation values are reported.
Back to article page