The Algorithm Research For Passive Localization In WLAN Based On Heuristic Probabilistic Neural Network

WLAN(Wireless Local Area Networks)indoor localization technique has become a research focus in recent years owing to its advantages of inexpensiveness,high accuracy and strong expansibility.Since general WLAN localization techniques are able to acquire a target's positions only if the target carries equipment for transmitting or receiving signal,it can't be useful in some occasions where targets are not allowed to carry the equipment or targets don't want to involve in the process.Hence how to locate a target on the basis of existing WLAN infrastructures under such circumstance has been another hotpot in the research of indoor localization.WLAN Device-free Passive Localization technique can still acquire a target's positions without the target carrying any equipment.With a parameter of signal or channel as medium,the technique can be achieved by exploring the correlation between target's location information and impact that's made by the target on signal propagation.Most of the existing WLAN Device-free Passive Localization techniques,which measure the impact made by a target on signal through the parameter of RSSI(Received Signal Strength Indication),are subject to RSSI's disadvantages of poor stability and poor bearing capacity of information,and consequently,they can't meet the demands of high accuracy and high stability.Meanwhile,indoor complex multipath effect brings a great challenge to accurate calculation of target's positions as it causes serious interference in properties of the correlation between locations and signal characteristic values.To solve the problem mentioned above,the algorithm in this thesis mines out the correlation between target's positions and characteristic parameters by utilizing the PNN(Probabilistic Neural Network),having taken advantage of CSI's(Channel State Information)prominent characteristics and restrained the interference caused by environment with the help of parallel subcarriers mechanism in OFDM(Orthogonal Frequency Division Multiplexing).Specifically,the main content of this thesis includes the three following parts.1.To detect the appearance of a target accurately,firstly this thesis makes a thorough inquiry about the impact made by the appearance of an unknown target on amplitude of CSI and then finds out that CSI shows differing patterns between when there is a target and no target.After that,the accurate detection of a target's appearance can be achieved through logical characteristic extraction and by means of non-parametric estimation.2.To reduce the interference in the process of target's position estimation made by multipath fading,this thesis proposes the data preprocessing mechanism to restrain environmental disturbance with the help of parallel subcarriers that have different carrier frequencies in OFDM,by filtering subcarriers according to the fading characteristics exhibited from their respective sub-channels.3.To properly construct a mapping model that represents how the characteristic values map to target's locations,this thesis introduces the pattern recognition method to solve the complex mapping problem between characteristics and locations.Since different multipath distributions will lead to alteration of characteristics' dimension,multiple mapping models are required to be trained in the off-line stage while the quick training method of PNN can avoid complexity of the algorithm.Besides,heuristic search algorithm is used to optimize parameters of PNN,not only improving the location accuracy of the algorithm but also making the algorithm more robust to environment's alteration.