SiMWiSense: Simultaneous Multi-Subject Activity Classification Through Wi-Fi Signals

 SiMWiSense uses Wi-Fi Channel State Information (CSI) and a few-shot learning algorithm to classify activities of multiple subjects simultaneously, overcoming scalability issues and enhancing generalization with minimal data.

Copy Image URL via instructions below

https://nu.testtechnologypublisher.com/files/sites/mark-saulich-.jpg

Background:

In the field of activity recognition, leveraging Wi-Fi signals presents a significant opportunity for monitoring individual activities within a given space. Traditional methods have primarily focused on classifying activities of a single human subject, limiting their effectiveness in environments with multiple subjects. This limitation poses a challenge for applications in surveillance, elder care, and smart homes, where robust multi-subject sensing is essential. Current approaches struggle to scale with increasing subject numbers, leading to exponentially growing classification challenges, and they exhibit poor performance in generalizing to new subjects or environments. This lack of adaptability and reliability across various settings without extensive reconfiguration or retraining hinders their practical real-world application.

Description:

Northeastern researchers have created SiMWiSense, a technology that uses Channel State Information (CSI) from Wi-Fi signals for classifying activities of multiple subjects simultaneously, overcoming the limitations of single-subject systems. This system efficiently handles the complexity of multiple subjects by using CSI data from the nearest device. Additionally, SiMWiSense employs a few-shot learning algorithm called Feature Reusable Embedding Learning (FREL), which improves its adaptability to different environments and subjects with minimal data, addressing the challenges faced by current methods. SiMWiSense has demonstrated its potential in enhancing activity recognition for applications in surveillance, elder care, and smart homes.

Benefits:

• Scalable for environments with multiple subjects
• Enhanced adaptability across various settings
• Reduced data requirements for effective learning
• Leverages existing Wi-Fi infrastructure
• Provides non-invasive monitoring, eliminating the need for wearable sensors

Applications:

• Smart home automation for resident activity monitoring
• Elderly care centers for behavioral analysis and fall detection
• Security systems for multi-subject surveillance
• Healthcare facilities for patient monitoring and movement analysis
• Retail analytics for customer flow and behavior analysis

Opportunity:

Research collaboration

licensing

Put in Links to Lab website and USPTO patent: