MULTIFRACTAL DISCRIMINATION MODEL (MDM) OF HIGH-FREQUENCY PUPIL DIAMETER MEASUREMENTS FOR HUMAN-COMPUTER INTERACTION - Quantitative Medical Data Analysis Using Mathematical Tools and Statistical Techniques - page 342

Multifractal spectrum

0.1

Right slope

Left slope

0

C f = −0.2

−0.1

−0.2

Broadness

−0.3

−0.4

−0.5

Mode

a 2

a 1

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8

a (q)

Fig. 4. Canonical features of the multifractal spectrum , including the left Slope,
Broadness, and spectral Mode measures.

Table 2. Summary statistics of the multifractal spectral characteristics for
our PRB datasets
Group Left Slope Spectral Mode Broadness

Control Mean 0.5053 0.4177 0.8591

Median 0.4725 0.4153 0.7668

Std. 0.1658 0.1517 0.4956

#1 Mean 0.3787 0.3561 0.8404

Median 0.3701 0.3214 0.7266

Std. 0.0738 0.1511 0.6796

#2 Mean 0.4049 0.4233 0.6989

Median 0.3908 0.4104 0.6804

Std. 0.1105 0.0985 0.1655

#3 Mean 0.484 0.3965 1.348

Median 0.4608 0.3926 0.8562

Std. 0.139 0.1723 1.1761

4. PRB Data Analysis

As mentioned previously, we attempt to nd the inherent features which
can separate the measurements with dierent ocular pathologies from each
other. The empirical evidence (e.g., wavelet-based energy spectrum) has
shown that PRB measurements possess self-similarities and fractalities.
Hence, it is natural to apply multifractal spectra to discriminate these mea-
surements.