Dual-Channel Speech Enhancement Using a Perceptual Filterbank for Hands-Free Communication - Digital Signal Processing for In-Vehicle Systems and Safety - page 178

Fig. 12.1 Schematic diagram of TFGSC

a schematic diagram of the dual-channel TFGSC is shown with the signal propaga-
tion model in the frequency domain.
The transfer function ratio H is defined by

A 2
A 1 :

H ¼ (12.1)

Through the FBF, primary signal is given by

¼

1 1

Z 1
Z 2

W C Z H N 2

1 H

Y ¼ ¼ ½ A 1 S þ

2 N 1 þ

½ : (12.2)

2

1 þ jj H 1 þ jj H

The FBF forms a beam in the look direction to pass speech and outputs a signal
consisting of the distortionless speech and noise components including both the
directional interference and the in-vehicle ambient noise. Next, BM forms a null
beam to block speech and produces the noise reference signal:

¼

Z 1
Z 2

B H Z U ¼ ¼ ½ H 1 HN 1 þ N 2 : (12.3)

The noise reference signal generated goes to an NC block, and it constructs a
filter, G B , to estimate and eliminate the noise component in FBF output via the
general wiener filter solution as [ 4 ]

H

¼ F UY

E ½ UY c

G B ¼

(12.4)

E ½ UU H

F UU

G B U

Y o ¼ Y : (12.5)