Combination of DNN and Improved KNN for Indoor Location Fingerprinting

Algorithm 1

Pseudocode of online positioning process.

Input: RSSI

Output: Target position (P)

 Get the collected RSSI vector R();

 Initialize distance set D as an empty set;

 Initialize weight set as an empty set;

 Initialize DNN model from the training file;

 A certain cluster C that R belongs to can be predicted by DNN model;

 for (each reference RSSI vector in C) do //RSSI fingerprint traversal

   Calculate the squared Euclidean distance between R and ;

   Add into D;

   Add the number of same APs () between R and into ;

 end for

 Select k nearest neighbors and their corresponding weights according to D;

 for (each position in k nearest neighbors) do // stands for coordinate of the selected position

   ; //P is the coordinate of the final position

 end for