Contactless Activity Monitoring for Pervasive and Ubiquitous Sensing 


Research Objective

We seek to develop innovative ultrasonic echolocation-based human sensing solutions for emerging healthcare applications, such as in-home physical rehabilitation support [1, 3], social interaction tracking for depression sufferers [5], sleep quality assessment [4], and sedentary activity tracking within office environments [2, 6]. By employing ultrasonic echolocation as a sensing mechanism, our systems offer unique value propositions relative to competitive approaches, based largely upon their capacity to preserve individual privacy, minimize computational complexity, and integrate commercial off-the-shelf (COTS) components and systems to minimize cost. Our systems leverage the demonstrated expertise of the NeEWS laboratory in the area of intelligent distributed systems to create cloud-enabled architectures suitable for large-scale development for a variety of applications. An example of such an end-to-end cyber-physical architecture is depicted in the figure below for applications of our technology to office activity tracking.


Office Activity System


Proposed Mechanism

As demonstrated in our recent publications, we perform human sensing by deploying emerging supervised learning algorithms on data generated from active pulsed COTS ultrasonic transducers integrated with our various system form-factors (both desktop array architectures as depicted above as well as single-sensor wearable badges as depicted below). In our preliminary work, classification is accomplished using only the measured distance between the sensors and human targets, hereby referred to as first-reflection echolocation, with distance computed using the traditional time-of-arrival approach employed in classical ranging systems. Such an approach presupposes the existence of distinguished kinematic signatures for the various activities of interest which may be captured within the temporal (ie: sampling frequency) and spatial (ie: target of interest is the most proximal within the radiation beam) constraints of the sensor-target geometry. While such constraints are inherently restrictive, they afford dramatically simplified computational frameworks using only time-series processing techniques.

To relax the constraints associated with first-reflection echolocation, our current work focuses on processing the full series of analog reflections associated with the sensing environment, hereby referred to as full-wave reflection processing. For a pulsed ultrasonic transducer architecture, such a signal consists of a superposition of reflections of various amplitudes and delays, which superimpose to create the full-wave signal of interest at the receiver output. We are currently investigating various techniques by which to compute features of interest which are most indicative of the underlying movements occurring withn the reflecting environment, along with optimal machine learning frameworks to support classification for the various applications of interest. 



Human Detection Badge


Selected Publications

[1] H. Griffith and S. Biswas, Home-Based Upper Extremity Rehabilitation Support Using a Contactless Ultrasonic Sensor, 39th International Conference of the IEEE Engineering in Medicine and Biology Society (EMBS), July 2017, Jeju Island, Korea. (Submitted)

[2] H. Griffith and S. Biswas, Office Activity Classification Using First-Reflection Ultrasonic Echolocation, 39th International Conference of the IEEE Engineering in Medicine and Biology Society (EMBS), July 2017, Jeju Island, Korea. (Submitted)

[3] H. Griffith, R. Ranganathan, and S. Biswas, Towards a First-Reflection Ultrasonic Sensor Array for Compensatory Movement Identification in Stroke Sufferers, 35th IEEE International Performance Computing and Communications Conference (IPCCC), December 2016, Las Vegas, NV.

[4] H. Griffith, Y. Shi, and S. Biswas, Contactless On-Bed Activity Sensing Using First-Reflection Echolocation, 38th International Conference of the IEEE Engineering in Medicine and Biology Society (EMBS), August 2016, Orlando, FL.

[5] H. Griffith, Y. Shi, and S. Biswas, A Wearable System for Asymmetric Contactless Human Sensing, 38th International Conference of the IEEE Engineering in Medicine and Biology Society (EMBS), August 2016, Orlando, FL.

[6] S. Biswas, et al. "Contact-less indoor activity analysis using first-reflection echolocation." IEEE International Conference on Communcations (ICC), E-Health Symposium, May 2016, Kulala Lumpur, Malaysia.