Published in


Samsung AI Uses WiFi Signals to Generate Consistent In-Home User Localization Data

An increasing number and variety of sensors and virtual assistants are being deployed in the places we live, laying the digital foundation for the smart home environment of tomorrow — where lighting, climate, entertainment systems, alarm systems, appliances and more are Internet-connected and automatically controlled.

The optimization of such smart home systems requires constant user location information with at least metre-level resolution. Home-based localization and activity monitoring based on 24-hour image or video-based sensors however is often and understandably resisted due to privacy concerns. Also, most current location sources like GPS are not available for use indoors. This has prompted researchers like Xi Chen to turned their eyes to the ubiquitous indoor WiFi signals.

In a recent paper, Chen and his colleagues with the Samsung AI Center utilized the WiFi signals to establish a submeter-level localization system that employs WiFi propagation characteristics as users’ location fingerprints. The researchers also propose a WiFi-based Domain-adaptive system (FiDo) , which is able to localize new users without labelling their data.

“Indoor localization systems like FiDo can be very useful for developing intrusion detection and fall detection devices, or to add new features such as presence and activity detection for existing smart devices,” Chen told Synced.

Existing WiFi-based localization systems can generally be divided into two categories: device-oriented and device-free. Device-oriented systems use the pre-measured locations of WiFi Access Points as references and then measure the real-time localization and movement information using another device carried by the user. Device-free systems on the other hand pinpoint a user’s physical presence by learning changes in WiFi propagation caused by their body to create WiFi location fingerprints.

Although device-free systems are more flexible and easy-to-deploy, WiFi location fingerprints generated by such systems may experience significant inconsistency across different users. A localization system trained on one user’s fingerprints is very likely inapplicable for another user with a different body shape.

“No customers would want to pay for such an inconsistent system, with which they have to adjust or reset every time there’s a change among different users or environments,” Chen explains. “And in reality, once deployed, it’s hard for a system itself to label all the new input data.”

To come up with a practical solution, the team developed an algorithm for FiDo that can learn new incoming data in an unsupervised manner and thus provide robust location information for different users. The researchers say the algorithm “achieves a perfect balance between classifying the labelled data and extracting domain-independent features from unlabelled data.”

An overview of FiDo

FiDo utilizes a data augmenter to generate synthetic fingerprints based on the collected WiFi location fingerprints, and Variational Autoencoders to summarize their statistical features. In this way, fingerprints collected from one or two example users could build a larger group of ‘virtual users’ which enable the system to more accurately localize new users.

FiDo also learns a domain-adaptive classifier based on a special neural network with a joint classification reconstruction structure. This enables it to predict locations in different domains, covering both WiFi fingerprints from labelled data of the example users and the newly collected unlabelled WiFi data from any new users.

To evaluate FiDo, the researchers established a submetre-level localization testbed, accurate to about 70 centimetres, using commercial off-the-shelf WiFi devices. They recorded the WiFi Channel State Information data from nine volunteers with different genders, heights, and weights; and at different locations.

Compared to the previous SOTA model, FiDo is able to improve average recall, precision, and F1 score by 11.7, 9.2, and 11.8 percent respectively. Compared to previous methods, it also largely boosted the worst-case True Positive Rates across different people.

Although the team achieved nearly 85 percent accuracy in both recall and precision rates, Chen says the team needs to improve system accuracy by about 10 percent before FiDo can be fully commercialized, and that this may require not only algorithm improvements but also adding other signal sources such as Bluetooth and sound waves.

The paper FiDo: Ubiquitous Fine-Grained WiFi-based Localization for Unlabelled Users via Domain Adaptation can be found here.

Journalist: Yuan Yuan | Editor: Michael Sarazen

Thinking of contributing to Synced Review? Synced’s new column Share My Research welcomes scholars to share their own research breakthroughs with global AI enthusiasts.

We know you don’t want to miss any story. Subscribe to our popular Synced Global AI Weekly to get weekly AI updates.

Need a comprehensive review of the past, present and future of modern AI research development? Trends of AI Technology Development Report is out!

2018 Fortune Global 500 Public Company AI Adaptivity Report is out!
Purchase a Kindle-formatted report on Amazon.
Apply for Insight Partner Program to get a complimentary full PDF report.



Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store


AI Technology & Industry Review — | Newsletter: | Share My Research | Twitter: @Synced_Global