Cultural Heritage Applications: Archaeological Ceramics and Building Restoration - On Statistical Pattern Recognition in Independent Component Analysis Mixture Modelling - page 148

Fig. 6.10 Power B-Scan after ICA preprocessing a B-Scan from sinusoidal components b B-
Scan from backscattered components

(a) (b)

1
2
3
4
5
6
7
8
9
10

0 0.5 1 1.5 2 2.5 3 3.5

x 10 -5 Time ( s)

Fig. 6.11 Recorded signals and centroid frequency B-Scan by non-stationary analysis

a Recorded signals b Centroid frequency B-Scan

controlled interference was added to the signals testing different frequencies and
amplitudes of the interference. Figure 6.12 a shows the error in the extraction of the
interference versus the ratio power interference to the power signal (Pinterference/
Psignal) for different interference frequencies. The higher the interference
amplitude, the better the extraction of the interference; the higher the interference
frequency, the worse the extraction of the interference. Figure 6.12 b depicts an
enhanced centroid frequency B-Scan (cancelled interference) with a layer that is
clearly defined at 4 mm. corresponding to the mortar layer of the dome wall.
Other alternatives for sinusoid extraction are based on the use of the so-called
notch filtering [ 34 ]. These alternatives can be designed assuming prior knowledge
of the pulsations to be cancelled. Thus, BSS could be used as a prior step to notch
filtering; however, it implies transient effects, possible instability problems, and
some distortion of the obtained interference-free records (because of the finite
notch bandwidth).