Speech Enhancement - Digital Speech: Coding for Low Bit Rate Communication Systems - page 417

2

0.5

h 92

0.45

0.4

1

0.35

h 85

0.3

0.25

0

h 20

0.2

0.15

h 35

0.1

− 1

0.05

h 0

0

− 2 − 0.05

80000 1 × 10 5 0 20000 40000 60000 80000 0 20000 40000 60000

Time (samples) Time (samples)

Figure 11.32 Performance of noise-echo canceller for clean speech

track have a considerable effect on the filter coefficient adaptation. Otherwise
the overall step size due to w k (i) (i.e. w k (i)β ) will be small, therefore not
changing the previous filter coefficient value by much and thus reducing the
likelihood of divergence.
A similar result was obtained in the second experiment when a more
complex echo was used. The echo used for this setup was generated through
the sum of three different delays: 20, 40 and 60 samples with corresponding
attenuation factors of 0.2, 0.48 and 0.35 respectively. Figures 11.32-11.37
show the results obtained for the second setup which proves the effectiveness
of the new adaptation algorithm proposed by Al-Naimi [24].

11.5 Summary

With advanced signal processing algorithms and techniques it is possible
to improve the quality of speech communications significantly. Both echo
and noise cancellation/suppression algorithms have been reasonably well
developed to tackle high levels of echo and noise present in communication
systems. It is, of course, important to adapt the existing algorithms to
specific communication systems to maximize their performances. When both
acoustic noise and echo are present it is important to tune the overall
enhancement algorithms (noise suppressor and echo canceller) jointly to
maximize performance. Another important issue is the convergence time of