Information Technology Reference
In-Depth Information
Figure 4. Typical ECG recording of a normal human heart
•
A registration phase in which the set of
ECG features and patterns of individual
subjects are extracted from their ECG re-
cording and digitized to constitute a collec-
tion of biometric templates. Several stages
of preprocessing and screening processes
may be applied to filter out bad quality
templates. Accepted templates are stored
in a central database to support the recog-
nition phase discussed below.
•
Soft independent modeling of class anal-
ogy (Esbensen et al., 2002)
•
Wavelet distance measurement (Chan et
al., 2008)
•
Mahalanobis distance (Mahalanobis, 1936)
A possible set of ECG features that can be
extracted and used in the automatic identifica-
tion process is presented in Figure 5. The work
in (Biel et al., 2001) introduces an extensive set
of thirty unique ECG features that can be used in
the classification mechanism.
In summary, an ECG identification system is
composed of three main components:
•
A recognition or identification phase in
which the ECG characteristics of the hu-
man subject to be identified are extracted
and statistically compared against the bio-
metric template captured in the registration
phase. A similarity score is derived be-
tween the extracted ECG features and the
stored templates. The template comparison
with the highest similarity score indicates
the identity of the individual.
1. The ECG electrodes or sensors attached to
the human chest and periphery and used to
capture the heart electrical activity.
2. The ECG feature extraction process with
which the required ECG features are ex-
tracted from the ECG recording.
3. The matching or classification process in
which the subject's ECG features are checked
against a set of templates to determine a
best match that indicates the identity of the
subject.
Some of the functions used in the literature for
establishing the similarity score are:
•
Mean-square deviation (Anderson &
Woessner, 1992)
Search WWH ::
Custom Search