LOS and NLOS Signal Classification based on Advanced Particle Swarm Optimization for Acoustic Self-Calibrating Indoor Localization

LOS and NLOS Signal Classification based on Advanced Particle Swarm Optimization for Acoustic Self-Calibrating Indoor Localization

연구 분야: Infrastructure

학회: ICIIP '19: Proceedings of the 4th International Conference on Intelligent Information Processing

The demand for indoor positioning is exploding. For indoor positioning, TPSN ranging model with specific frequency band is often used based on android platform or others. Time of arrival (TOA) is often used as an important part to achieve efficient localization. The signal wav files classification of line of sight (LOS) and non-line of sight (NLOS) will be involved. Support Vector Machine (SVM) is often used to complete classification in most positioning systems and the effect is not so good. At this moment, in order to improve the performance of this system, the Shrinkage Enhanced Particle Swarm Optimization (SEPSO) is introduced to improve the effect. Compared with traditional PSO, it introduces contraction coefficient to cope with the problem of few samples and non-linear system better than before. It also combines the non-linear decreasing inertia weight of asynchronous linear learning method with original algorithm to improve the ability of jumping out of local minimum area. Considering the actual use of those platforms, it improves the feature extraction algorithm and uses SEPSO to upgrade the performance of SVM from three aspects: training sample re-selection, parameter optimization and optimal preservation strategy (elitist strategy) of each generation population.