3D pedestrian dead reckoning and activity classification using waist-mounted inertial measurement unit

Abstract

In this paper, an algorithm to estimate the position of a pedestrian in a 3-dimensional space is introduced. The proposed algorithm exploits the data provided by a waist-worn inertial platform and does not rely on the presence of any external infrastructure. Relevant features are extracted from the accelerometer data and are used to detect pedestrian activities such as standing, walking, going upstairs, or going downstairs. The estimate of the position is updated through a step detection procedure, which combines the signals provided by the inertial platform with the information about the pedestrian activity class

Multi-Floor Indoor Pedestrian Dead Reckoning with a Backtracking Particle Filter and Viterbi-Based Floor Number Detection

“…It consisted of a combination of floor number detection with WiFi RSS MBF and floor transition detection using the accelerometer and barometer. In contrast to many comparable systems [ 23 , 25 , 28 , 29 , 55 , 56 ], our system detected both stairs and elevator usage, and these detections served a dual purpose: to aid in determining the sequence of visited floors ( Section 3.3 ) and to improve 2D localisation ( Section 3.4.3 ).…”

“…Floor transition detection using machine learning and features extracted from multiple sensors (accelerometer, gyroscope and/or barometer) was successfully implemented in [ 19 , 24 , 28 , 50 ]. References [ 28 , 29 ] obtained over 90% accuracy in distinguishing between going upstairs and downstairs, but elevators were not detected. References [ 24 , 50 ] detected both stairs and elevator usage, but it performed worse at distinguishing the direction of stairs usage.…”

“…The cumulative height change during a transition can then be used to estimate the amount of changed floors. Height change was accurately estimated in [ 29 , 51 ] using only the IMU sensors, but this was achieved using dedicated strapdown IMUs, which provide higher accuracy than unconstrained smartphones [ 20 ]. The barometer was used successfully as an alternative to detect floor transitions by converting the measured pressure to height [ 52 ].…”

“…Another challenge in an indoor pedestrian localisation system is detecting floor transitions and determining the correct floor number (i.e., floor detection), thus allowing a pedestrian to be tracked across multiple floors. This can be achieved by detecting the floor number directly using (WiFi) fingerprinting [ 25 , 27 ] or by detecting floor transition events using data from onboard sensors [ 24 , 28 , 29 , 30 , 31 ].…”

Pedestrian Dead Reckoning With Wearable Sensors: A Systematic Review

“…Studies with a single IMU, most frequently mounted the sensor on one of the shoes [33]. Another commonly selected placement was the waist [34]. Studies with multiple IMUs preferred to mount them on the lower limb and trunk [35].…”