Although WiFi fingerprint based localization can achieve better accuracy than triangulation, it still
suffers from accuracy
uctuation problem. As a result, many filtering techniques have been developed
to mitigate the
uctuation of localization accuracy mainly focused on the post-processing stage of a
localization engine. Kalman filter, delay filter, map matching are such filtering techniques adopted and
used in both indoor and outdoor environments. However little work has been found for the pre-processing
stage of a localization engine. But in fact the pre-processing such as imputing missing signals of a captured
WiFi fingerprint can bring us as much effect as we can get through post processing performed by location
filters.
In WiFi fingerprint based localization, a WiFi radio map (WRM) is constructed in advance and
then a location for a captured WiFi fingerprint is estimated based on the WRM. Since we usually
collect WiFi fingerprints many times at a location, and then compute their average WiFi fingerprint
for the construction of WRM, almost all of the WiFi signals accessible at the location are included in
the constructed WRM. On the other hand, a WiFi fingerprint captured at a location in online phase for
localization inevitably has missing WiFi signals because it is obtained by scanning WiFi signals just once
or twice. Thus, if we estimate a location with the captured fingerprint without padding or imputation
of the missing signal, we cannot expect to achieve high localization accuracy.
In this paper we propose a technique to enhance the accuracy of WiFi-fingerprint based localization
by imputing missing signals of WiFi fingerprints. There can be many different ways of imputing or
handling missing signals. For example, we can use a predefined value instead for missing signals or we
can even disregard the missing signals for localization. In this paper we develop two techniques for this;
one is to impute missing signals by referring to the WRM fingerprint at the very previo...