An Adaptive Weighted KNN Positioning Method Based on Omnidirectional Fingerprint Database and Twice Affinity...

Area of Science:

• Indoor positioning systems
• Wireless communication
• Signal processing

Background:

• Human body significantly impacts Wi-Fi signal power, causing localization errors in traditional K-nearest neighbor (KNN) algorithms.
• Fixed K values in KNN do not account for environmental variations and signal obstructions.

Purpose of the Study:

• To develop an adaptive weighted KNN positioning method for enhanced Wi-Fi localization accuracy.
• To address the signal sheltering impact of the human body on Wi-Fi signals.
• To improve the performance of fingerprinting-based localization techniques.

Main Methods:

• Proposed an omnidirectional fingerprint database (ODFD) incorporating position, orientation, and mean received signal strength (RSS) sequence.
• Introduced affinity propagation clustering (APC) in the offline stage, fusing signal-domain and position-domain distances.
• Developed an adaptive weighted KNN algorithm using APC for online user positioning, clustering initial RPs and selecting the most probable sub-cluster.

Main Results:

• Achieved a mean positioning error of 2.2 meters.
• Obtained a root mean square error (RMSE) of 1.5 meters.
• Demonstrated superior performance compared to traditional fingerprinting methods.

Conclusions:

• The proposed adaptive weighted KNN method effectively mitigates human body interference in Wi-Fi signal power.
• The integration of ODFD and APC significantly enhances localization accuracy.
• This method offers a robust and accurate solution for indoor positioning challenges.