On the other hand, one of the distinct problems in industrial
safety is the acceptance by operators of the device performance.
Their acceptance can be very dependent on alarm effectiveness.
For example, a high number of false alarms will cause workers to
ignore or respond slowly to a real emergency. Swets et al. (2000)
address this problem through statistical prediction rules and what
are called receiver operating characteristic curves (ROC). The ROC
curves can be used to evaluate the tradeoff (in terms of probability)
between true and false alarms of different algorithms. ROC analysis
can help to choose a specific threshold to activate an alarm, for
example, an RSSI value to make a decision on the activation of an
emergency stop. Strict thresholds limit false positives at the cost
of missing true alarms whereas lenient thresholds maximize true
alarms at the cost of many false alarms. Which threshold is optimal
depends on several factors such as the cost of false alarms and the
seriousness of the hazards. The aforementioned research related to
location distinction and perimeter distinction use ROC curves as a
quantitative criterion to compare the performance of the different
algorithms proposed. Location distinction and perimeter distinction
methods are complementary to the existing localization techniques
and could also be applied for industrial safety purposes.
The protective equipment proposed here can be classified as an
electro-sensitive safety barrier with the capacity to distinguish
whether the object entering a dangerous area is a workpiece or
the body part of a worker (to which a transmitter is attached).
Although there are no standards specific to radio frequency presence
sensing devices (OSHA, 1987), there exist regulations for general
requirements and tests for protective devices; examples
include EN 999 (1998) or EN 61496-1 (2004). Regulations also exist
for the application of protective devices to specific machinery, e.g.,