Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Suite-IN: Aggregating Motion Features from Apple Suite for Robust Inertial Navigation

Lan Sun, Songpengcheng Xia, Junyuan Deng, Jiarui Yang, Zengyuan Lai, Qi Wu, Ling Pei

TL;DR

This work proposes a multi-device deep learning framework named Suite-IN, aggregating motion data from Apple Suite for inertial navigation, and innovatively introduces a contrastive learning module to disentangle motionshared and motion-private latent representations, enhancing positioning accuracy.

Abstract

With the rapid development of wearable technology, devices like smartphones, smartwatches, and headphones equipped with IMUs have become essential for applications such as pedestrian positioning. However, traditional pedestrian dead reckoning (PDR) methods struggle with diverse motion patterns, while recent data-driven approaches, though improving accuracy, often lack robustness due to reliance on a single device.In our work, we attempt to enhance the positioning performance using the low-cost commodity IMUs embedded in the wearable devices. We propose a multi-device deep learning framework named Suite-IN, aggregating motion data from Apple Suite for inertial navigation. Motion data captured by sensors on different body parts contains both local and global motion information, making it essential to reduce the negative effects of localized movements and extract global motion representations from multiple devices.

Suite-IN: Aggregating Motion Features from Apple Suite for Robust Inertial Navigation

TL;DR

This work proposes a multi-device deep learning framework named Suite-IN, aggregating motion data from Apple Suite for inertial navigation, and innovatively introduces a contrastive learning module to disentangle motionshared and motion-private latent representations, enhancing positioning accuracy.

Abstract

With the rapid development of wearable technology, devices like smartphones, smartwatches, and headphones equipped with IMUs have become essential for applications such as pedestrian positioning. However, traditional pedestrian dead reckoning (PDR) methods struggle with diverse motion patterns, while recent data-driven approaches, though improving accuracy, often lack robustness due to reliance on a single device.In our work, we attempt to enhance the positioning performance using the low-cost commodity IMUs embedded in the wearable devices. We propose a multi-device deep learning framework named Suite-IN, aggregating motion data from Apple Suite for inertial navigation. Motion data captured by sensors on different body parts contains both local and global motion information, making it essential to reduce the negative effects of localized movements and extract global motion representations from multiple devices.

Paper Structure

This paper contains 17 sections, 6 equations, 5 figures, 4 tables.

Table of Contents

  1. INTRODUCTION
  2. Related Work
  3. Data-driven Pedestrian Localization
  4. Contrastive Learning for Wearable sensors
  5. Dataset Description
  6. Dataset overview
  7. Data Collection Apparatus and Process
  8. Method
  9. Problem Statement
  10. Motion-shared Representation Learning for pedestrian localization estimation
  11. Motion-aware Contrastive Module for Share- and Private-motion Learning
  12. Overall Network Parameters Optimization
  13. Experiment
  14. Experimental Setup
  15. Comparison with the State-of-the-Art
  16. ...and 2 more sections

Figures (5)

  • Figure 1: Our innovative data-driven approach for robust pedestrian localization: achieving robust positioning under complex walking patterns and flexible equipment configurations by extracting global motion information.
  • Figure 2: Illustration of device wearing when we collect data
  • Figure 3: Overview of our multi-device fusion positioning framework. Our proposed model comprises two technical modules: 1) motion-shared representation learning and 2) motion-aware contrastive learning module.
  • Figure 4: Selected visualizations. We select 2 examples from each walking range and visualize the reconstructed trajectories of competing methods. For each sequence, we label the trajectory length, ATE and RTE of our method and the second-best method.
  • Figure 5: Position Estimation Error. Left: the position estimation error of the sequence where the watch is taken off midway. Middle: the sequence where the headphones are taken off midway. Right: the sequence where three devices are randomly shaken.