چكيده به لاتين
Business tendency towards indoor positioning services is going through an upward trend,
resulting in development of many indoor positioning methods. Lack of accurate indoor
positioning in many cases has resulted in attention to be paid to other methods. Also, from
among signals, WiFi signal has been welcomed due to comprehensive deployment.
Recently, many techniques have been developed in the field of WLAN positioning;
however, WiFi fingerprint-based indoor positioning has been specifically taken into
consideration. The reason is that it does not require lines of sight of access points to be
measured; and, it can have numerous applied capabilities in complex indoor environments.
In general, proposed positioning methods are divided into several categories. One of these
approaches is sparse positioning with lower positioning error, compared to other positioning
methods. However, implementing analyzing method of sparse signals has its own problems.
Main problem with sparse positioning methods is that implementing takes place with low
processing speed. To solve the problem, either equipments with higher processing speed are
required which would be resulted in higher cost; or, computations have to be made simpler.
In general, in statistical modeling of sparse signals, those functions with high capability of
showing one single sparse signal would be used. These functions have common
characteristics including having heavy tail and sharp-pointed peak at source. In the thesis
and considering the grouping made based on these characteristic, new models based on
Bayesian compressive sensing have been presented, to the aim of increasing processing
speed. Considering the simulation performed, it could be observed that positioning
processing speed has been considerably increased; meanwhile, almost no change has been
made in terms of positioning error.
Keywords: Statistical (Bayesian) compressive sensing; fingerprint-base indoor
positioning; indoor sparse positioning methods
