In this section, we demonstrate the effect of the number
of PIR sensors on recognition accuracy in classifying walking
directions. Table I summarizes the average of 10 times 10-fold
cross-validation results over the selected classification methods
based on PIR signals. We do not show standard deviation of
the results because the average values are very close to each
other (i.e., very small standard deviation) for all classifiers.
In Table I, we can know that instance-based learning
method (k-nearest neighbor algorithm) shows the best