A Pose Awareness Solution for Estimating Pedestrian Walking Speed

• Published in: Remote sensing (Basel, Switzerland) Vol. 11; no. 1; p. 55
• Main Authors: Guo, Guangyi, Chen, Ruizhi, Ye, Feng, Chen, Liang, Pan, Yuanjin, Liu, Mengyun, Cao, Zhipeng
• Format: Journal Article
• Language: English
• Published: Basel MDPI AG 01.01.2019
• Subjects: Accelerometers; Accuracy; Algorithms; Architectural engineering; Artificial intelligence; Cellular telephones; Estimation; Feature extraction; Field tests; International conferences; Learning algorithms; Machine learning; multi-sensors; Neural networks; Pedestrians; posture context; Radio frequency; Remote sensing; Sensors; Short message service; smartphone; Smartphones; Ubiquitous computing; Walking; walking speed estimation; United States > US; China
• ISSN: 2072-4292; 2072-4292
• DOI: 10.3390/rs11010055

Pedestrian walking speeds (PWS) can be used as a “body speedometer” to reveal health status information of pedestrians and positioning indoors with other locating methods. This paper proposes a pose awareness solution for estimating pedestrian walking speeds using the sensors built in smartphones. The smartphone usage pose is identified by using a machine learning approach based on data from multiple sensors. The data are then coupled tightly with an adaptive step detection solution to estimate the pedestrian walking speed. Field tests were carried out to verify the advantages of the proposed algorithms compared to existing solutions. The test results demonstrated that the features extracted from the data of the smartphone built-in sensors clearly reveal the characteristics of the pose pattern, with overall accuracy of 98.85% and a kappa statistic of 98.46%. The proposed walking speed estimation solution, running in real-time on a commercial smartphone, performed well, with a mean absolute error of 0.061 m/s, under a challenging walking process combining various usage poses including texting, calling, swinging, and in-pocket modes.