A Smartphone Camera-Based Indoor Positioning Algorithm of Crowded Scenarios with the Assistance of Deep CNN

Area of Science:

• Electrical Engineering
• Computer Science
• Signal Processing

Background:

• Received Signal Strength Indicator (RSSI)-based indoor positioning is common but struggles with accuracy in crowded environments due to signal interference.
• Existing RSSI methods do not adequately address the impact of human presence on signal strength.

Purpose of the Study:

• To develop an improved wireless signal compensation model for indoor positioning systems.
• To enhance signal strength estimation accuracy by incorporating population density, distance, and frequency.

Main Methods:

• A convolutional neural network (CNN)-based human detection approach was used to estimate the number of individuals.
• A novel model was developed to describe the relationship between population density and signal attenuation.
• Trilateration was employed for pedestrian localization using the compensated signal strength data.

Main Results:

• The proposed model increased signal strength estimation accuracy by 1.53 times compared to models that ignore human presence.
• Achieved a localization accuracy of less than 1.37 meters in crowded indoor scenarios.
• Demonstrated superior performance over other existing RSSI-based indoor positioning models.

Conclusions:

• The developed wireless signal compensation model effectively mitigates interference caused by population density.
• The proposed method significantly improves indoor positioning accuracy and reliability.
• This approach offers a viable solution for enhancing location-based services in crowded indoor settings.