Transportation mode detection using cumulative acoustic sensing and analysis

  • PDF / 290,019 Bytes
  • 3 Pages / 612.284 x 802.205 pts Page_size
  • 0 Downloads / 183 Views

DOWNLOAD

REPORT


Transportation mode detection using cumulative acoustic sensing and analysis Dinesh VIJ

, Naveen AGGARWAL

University Institute of Engineering and Technology, Panjab University, Chandigarh 160014, India c Higher Education Press 2020 

1

Introduction

Smartphone sensing has emerged as an active area of interest for researchers, as these are increasingly being used in smart transportation services. This paper focuses on a smartphone based vehicle monitoring system, viz. transportation mode detection. Detection of the transportation mode of a commuter is usually carried out through a combination of motion sensors such as an accelerometer and position sensors such as GPS [1]. In the recent years, few researchers have presented the usage of single smartphone sensors like accelerometer [2] that consume less energy for detecting the transportation modes. However, these techniques work by detecting the movement and analyzing the speed of the commuter for a certain period of time. But the use of speed estimates is not reliable, because speed of the commuter might depend on various aspects such as the traffic state. So, different vehicles might generate similar speed signature under specific conditions. Another problem with most of these techniques is the requirement of a substantial amount of time to infer the final vehicle category label [3,4]. Further, most of these algorithms work in two phases. In the first phase, the algorithm detects when the user switches from one transportation mode to another. In the second phase, it then classifies the identified switch and gives it a label. Therefore, the transportation mode detection accuracy is constrained by the accuracy of mode-switch detection [5]. In 2014, Garg et al. [6] proposed a detection approach using a combination of accelerometer, gyroscope, orientation, GPS, magnetometer, light, and microphone sensors that needed 5 minutes of continuous smartphone sensing (average detection accuracy: 92.88%) to infer four vehicle classes namely bike, auto-rickshaw, bus, and car. Regarding microphone sensing, they worked on the recognition of one specific kind of sound signal recorded by the smartphone, i.e., ambient vehicle noise, for inferring the mode of transportation. But, they reported that the use of a specific vehicle noise lone does not give a good estimation for inferring the vehicle type and discouraged the use of microphone sensor. In 2018, Vij and Aggarwal [7] proposed the use of smartphone based cumulative acoustic sensing rather than a specific kind of acoustic noise. Though, they did so for traffic state detection instead of Received June 3, 2019; accepted October 11, 2019 E-mail: [email protected]

transportation mode detection. In this paper, we have carried forward the work done by Garg et al. [6] and Vij and Aggarwal [7] and proposed to use the ubiquitously-available low-cost microphone sensor of the commuter’s smartphone and cumulative acoustic sensing for determining the mode of transportation. It needs only 30 seconds of cumulative acoustic sensing to predict