POF Sensing Grid: Signal Analysis to extract Human Motion Characteristics

This 2016 paper presents ways to analyze the signals of a low-cost plastic-optical-fiber sensing grid built into a carpet or mat. The methods track position, measure balance, and classify types of walking.

Presented at the 25th International Conference on Plastic Optical Fibres (POF2016) in September 2016, the paper was authored by Patricia Scully, Omar Costilla Reyes, Nabeel Ahmed, John Vaughan, Emma Stanmore, and Krikor Ozanyan of the University of Manchester. The University of Manchester’s research portal lists the conference as held at Aston University in Birmingham. The paper follows the group’s iMagiMat paper at the previous year’s conference, which introduced a tomography-based plastic optical fiber (POF) sensing layer for a smart carpet, and turns to a different question: what can be extracted from the grid’s raw responses once it is working.

The sensing grid uses low-cost, standard polymer optical fiber built into a standard carpet or rubber mat. Because pressing or bending a fiber changes how much light it transmits, the mat can collect data unobtrusively and over long periods from people walking or exercising on it. The paper presents three families of analysis methods. Tomographic imaging, which reconstructs an image from many fiber measurements, detects a person’s position, path, and movement anywhere on the mat. Analog null-balance methods measure the balance of a person standing still and how their center of mass shifts. Pattern recognition and machine learning identify different types of walking, the strand that connects to the machine-learning methods Dr. Costilla Reyes developed in his doctoral research.

Gait and balance are widely used indicators of mobility and health, and measuring them usually requires a clinic visit or a wearable device. A floor that records position, balance and walking type as people go about ordinary activities offers a way to gather this information unobtrusively and at low cost, which suits assessment in care and rehabilitation settings. By setting out what the grid can measure and the families of algorithms that read it, the paper gives the sensing platform a clear basis for developing and evaluating further methods. The same group’s gait-classification work with the carpet appears in Dr. Costilla Reyes’s IEEE SENSORS papers and in his University of Manchester PhD thesis.