Indoor positioning technique which gets indoor human's or object's location information has a broad application future in military and civil domains. Indoor positioning can be realizied based on the principle of different intensity signal received from different places. Because that the variation of signal intensity can be forecasted, user position can be got even with a high accuacy according to the signal intensity.In this paper, some related conception, system, technique, algorithm and characteristic of wireless communication channel of wireless positioning are introduced; some related work and research direction of indoor positioning are analysed; and the methodology of indoor positioning grounded on signal intensity and BP nueral network is discussed detailly.The paper builds a indoor positioning model from three-layer BP nueral network which utilizes signal intensity as its input vector. The BP nueral network including input layer, hidden layer and output layer adopts parallel network architecture. The output signal of hidden layer is transimited to output layer, througth the output layer the result is outputed. In the forward transimitting process, input signal is transimited from imput layer, processed by hidden layer and then sent to ouput layer. The state of the nerve cell in next layer only can be influenced by that of last layer. If the output result is not a expected one, then go to propagation reversly, that is to say, returning error signal back through former connection channel. The mean square error can be minimised by modifying the weight of nerve cell in each layer. The newly proposed model is simulated in WLAN evironment in terms of IEEE 802.11b standard, and the experiment result shows that the model is effective on improving positioning accuracy.With the ultimate purpose of providing services for uses, the paper constructs a position service model, designs an architecture of server end and mobile device end in the model and describes some detail applications utilizing the architecture. |