Application to biomedical signals - Practical Biomedical Signal Analysis Using MATLAB - page 228

4.63 comparison of TOAE decomposition by means of enriched and Gabor dictionar-
ies is shown. One can see that in some cases two Gabor atoms are needed to describe
one component. The enriched dictionary provides a more sparse representation. Fur-
thermore, the new approach gets rid of the “pre echo” effect, i.e., presence of false
energy traces before the start of the signal visible on TF maps created from MP with
Gabor dictionary. Analyses of simulated and real signals demonstrated that enriched
dictionary provides correct estimation of latencies for components of short and long
decay time.
FIGURE 4.63: (SEE COLOR INSERT) Time-frequency distributions of OAE
signal obtained by means of MP with dictionary containing asymmetric functions
(left) and dictionary of Gabor atoms only (right). At the far right, Fourier spectrum
of the signal is shown.
A hybrid matching pursuit algorithm that included Fourier spectral information
was developed in [Notaro et al., 2007] with the aim of speeding-up computation
times. Additionally the procedure was capable of identification of atoms whose
latency-frequency relation was not compatible with the frequency-latency depen-
dence. These atoms could be associated with several known phenomena, either in-
trinsic, such as intermodulation distortion, spontaneous emissions, and multiple in-
ternal reflections, or extrinsic, such as instrumental noise, linear ringing, and the