Intelligent RSSI Signal Preprocessing Method Based on Fusion Filtering and Feature Enhancement

In indoor positioning systems utilizing Bluetooth Low Energy (BLE), the Received Signal Strength Indicator (RSSI) is highly susceptible to noise, multipath effects, and hardware heterogeneity, which degrade localization accuracy. To address these challenges, this paper proposes an intelligent RSSI signal preprocessing method. The method integrates adaptive multi-filter fusion and spatio-temporal feature enhancement to improve signal quality. First, an adaptive multi-filter denoising strategy is designed, combining Kalman, median, and Gaussian filters. The weights of these filters are dynamically adjusted based on real-time local variance evaluation, enabling robust noise suppression while preserving critical signal features. Second, a spatio-temporal feature enhancement module is introduced, leveraging a Temporal Convolutional Network (TCN) to extract temporal dependencies and embedding spatial layout information of BLE devices to enhance feature representation. Experimental results show that the proposed method significantly reduces the signal variance (from 98.95 dBm<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> to 10.80 dBm<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>) and improves the smoothness of the processed signals by 89.08%, demonstrating its superior performance in RSSI signal preprocessing. This approach offers enhanced signal stability and accuracy, providing more reliable data for subsequent positioning algorithms.