Accurate On-Body Distance Estimation using BLE RSSI and IMU Sensor Fusion

Abstract

It is useful to estimate on-body distances for human motion tracking and rehabilitation monitoring. We perform accurate short-range (<1 m) distance estimation by fusing BLE RSSI with Inertial Measurement Unit (IMU) orientation data. We developed a custom wearable system using ESP32 microcontrollers and IMUs and conducted a study with N=6 participants performing a series of defned arm movements. We trained a Long Short-Term Memory (LSTM) model, termed QuaternionRNN, which leverages temporal sequences of normalized quaternions and RSSI to predict the distance between sensors on the waist and wrist. The model achieved a median absolute distance error of ~3 cm on a generalized dataset, and between 4-7 cm in subject-wise cross-validation. We compared this to Unscented Kalman Filters with RSSI alone and RSSI plus orientation, and an RSSI-only LSTM, and found QuaternionRNN to have 2-4x lower errors. This work establishes a cost-effective and accurate solution for short-range non-line-of-sight distance sensing.