ITEM METADATA RECORD
Title: Footstep localization by acoustic signals
Authors: Dekkers, Gert
Van Den Broeck, Bert
Karsmakers, Peter
Vanrumste, Bart
Issue Date: 28-Mar-2014
Conference: ECUMICT 2014 location:Ghent, Belgium date:27-28 March 2014
Abstract: Containing healthcare costs related to the aging population is a well-known challenge for Western societies. Technological research is directed to enable older persons to stay longer and safer in their own dwelling. A possible solution is an audio based home monitoring system. Previous research indicated a relation between a person's gait and his or her health. This work focuses on localizing footsteps by use of acoustic information. The acoustic information is retrieved from a Wireless Acoustic Sensor Network (WASN) containing multiple linear microphone arrays (nodes). In the proposed WASN setup the nodes were uniformly distributed in a room and placed at ground level which makes it a two-dimensional localization problem. Processing is done in three differrent stages. At first footstep activity is detected by an energy based approach. Then sound energy coming from a certain direction is estimated for each node with Steered Response Power (SRP) combined with a PHAse Transform weighting. In the last stage the directional patterns of each node are combined to construct a two-dimensional localization map by using Least Squares (LS) or Global Coherence Field (GCF). Real life data has been collected in an office setting including two test subjects walking a predefined path of eight steps. Each person repeated the experiment eight times yielding a dataset with 128 footsteps. The proposed system achieved a smallest median of error of 31 cm for the best step of the eight predefined steps.
Publication status: published
KU Leuven publication type: IMa
Appears in Collections:Electrical Engineering (ESAT) TC, Technology Campus Geel
Technologiecluster ESAT Elektrotechnische Engineering

Files in This Item:
File Description Status SizeFormat
ECUMICT_gert.pdf Published 32KbAdobe PDFView/Open

 


All items in Lirias are protected by copyright, with all rights reserved.