Device-Free Human Pose Estimation Method Using Wi-Fi Signals

Opportunity

AdaPose enables cross-site device-free human pose estimation using commodity Wi-Fi, presenting many applications and benefits. It provides a natural and convenient way for users to interact with smart devices and appliances using gestures and poses, without wearing sensors or installing cameras. Users can control lights, adjust temperatures, or play music by simply moving in front of a Wi-Fi router. AdaPose helps users create realistic and personalised avatars for the metaverse. Users can capture their poses with their Wi-Fi devices and transfer them to their avatars, enhancing immersion and expression in virtual worlds without needing specialised equipment. AdaPose monitors and analyses poses and movements for health and fitness purposes, such as detecting falls, tracking exercises, or providing feedback. This improves the quality of life and well-being of users without compromising privacy or comfort.

Technology

AdaPose is based on Wi-Fi Channel State Information (CSI), measures of the wireless signals between the transmitter and the receiver. Wi-Fi CSI captures signal changes caused by human movements, such as reflection, attenuation, and scattering. By processing and analysing the Wi-Fi CSI data, AdaPose estimates human poses, including joint locations and angles.

AdaPose consists of two main components: a pose estimation model and a domain adaptation module. The pose estimation model is a deep neural network that takes Wi-Fi CSI as input and outputs pose parameters. Meanwhile, the domain adaptation module is a feature alignment algorithm that reduces domain shift between the source domain (where the model is trained) and the target domain (where the model is applied).

It leverages the consistency between the input and output of the pose estimation model, aligning the features of the source and target domains in a common latent space. This improves the accuracy and robustness of the pose estimation model across different scenes, enabling cross-site device-free human pose estimation with commodity Wi-Fi.

Applications & Advantages

• Provides a natural and convenient way for users to interact with smart devices and appliances using gestures and poses, without wearing sensors or installing cameras.
• Helps users create realistic and personalised avatars for the metaverse.
• Monitors and analyses poses and movements for health and fitness purposes, such as detecting falls, tracking exercises, or providing feedback.
• Improves the accuracy and robustness of Wi-Fi-based pose estimation across different scenes, enabling cross-site device-free human pose estimation with commodity Wi-Fi.
• Protects the privacy and comfort of users, as it does not require any visual or physical contact.