Human motion feature extraction in personal emergency response systems and methods

Abstract

A non-wearable Personal Emergency Response System (PERS) architecture is provided, having a synthetic aperture antenna based RF interferometer followed by two-stage human state classifier and abnormal states pattern recognition. Systems and methods transmit ultra-wide band radio frequency signals at, and receive echo signals from, the environment, process the received echo signals to yield a range-bin-based slow signal that is spatio-temporally characterized over multiple spatial range bins and multiple temporal sub-frames, respectively, and derive from the slow signal a Doppler signature and a range-time energy signature as motion characteristics of human(s) in the environment and optionally also derive location data as movement characteristics thereof. The decision process is carried out based on the instantaneous human state (local decision) followed by abnormal states patterns recognition (global decision).