Page 271 -

6.3 Syntactic Analvsis 259

Figure 6.11. FHR signal parsing after approximation by line segments. Line
segments are labelled as: h=green; u=red; d=cyan; U=magenta; D=blue. Final
states are represented in black. The original signal (in the background) is in grey.

We see therefore that a finite-state machine constitutes a parser for a regular
grammar. As a matter of fact, given the expressiveness of a state-diagram, one
usually goes directly to the specification of the state-diagram or the state-transition
table, omitting the specification of the grammar.
The SigParse program allows one to specify a state-transition table for
recognizing regular string grammars. We exemplify its use by considering foetal
heart rate (FHR) signals, such as the one depicted in Figure 6.1 1.
Imagine that we want to detect the presence of downward spikes (class
DSPIKE) in such signals. We start by performing a line segment approximation
usmg the Chebychev norm with a tolerance E,,,,,=3, and use slope thresholds of 0.3
and 20 for the segmenl labelling. These are colour-coded as shown in Figure 6.1 1.
Next, we use the following simple set of rules for the recognition of DSPlKE
strings:

For the detection of accelerations (class ACEL), the set of rules is somewhat
more elaborate:

{ACELeuA; AH^; AeuA; AH~B; BHuA; Bed). (6-12b)

This corresponds to the finite-state machine of Figure 6.12.
Events detected as "accelerations" by this machine may no1 qualify as true
accelerations in the clinical sense. They only qualify as accelerations if they satisfy
a set of conditions on the minimum duration and minimum amplitude. For