Radio Tomographic Imaging-Based Fall Detection

Amongst older adults, falls at home can be deadly. Fall detection systems are an important tool to quickly summon help in the event of a fall. Wearable devices such as help buttons have been a popular approach to the problem; however, studies show that they have a low efficacy in the event of a fall because they require the user to actively remember to use the device. In response to this issue, another branch of fall detection design has focused on environmental fall detection methods. Our literature research concluded that radio tomographic imaging was the best approach to fall detection based on criteria of reliability, user experience, and privacy concerns. In this project we implement radio tomographic imaging using ten National Instruments Universal Software Peripheral Radios (NI-USRPs). Each USRP transmits a pure tone using amplitude modulation. The full duplex capabilities of the NI-USRP 2920s enable baseband frequency multiplexing, allowing for simultaneous transmit and receive. We measure received signal strength across each directional link between USRPs and implement kernel-based radio tomographic imaging to determine the relative attenuation in each pixel. Future work for this project would focus on using particle filtering for a track-before-detect approach to fall detection.