Convolutional Neural Network-Based Travel Mode Recognition Based on Multiple Smartphone Sensors

Nowadays, large-scale human mobility has led to increasingly severe traffic congestion in cities, how to accurately identify people’s travel mode has become particularly important for urban traffic planning and management. However, traditional methods are based on telephone interviews or questionnai...

Full description

Saved in:
Bibliographic Details
Published in:Applied sciences 2022-07, Vol.12 (13), p.6511
Main Authors: Guo, Lin, Huang, Jincai, Ma, Wei, Sun, Longzhi, Zhou, Lianjie, Pan, Jianping, Yang, Wentao
Format: Article
Language:English
Subjects:
Accelerometers
Algorithms
Bicycles
Classification
CNN
Data collection
Deep learning
Global positioning systems
GPS
Gyroscopes
Methods
Mobility
Neural networks
Pattern recognition
Performance evaluation
Sensors
Smart cities
smartphone data
Smartphones
social sensing
Traffic congestion
Traffic management
Traffic planning
Transportation networks
Transportation services
Travel
travel mode
Travel modes
Weight reduction
Citations: Items that this one cites
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Nowadays, large-scale human mobility has led to increasingly severe traffic congestion in cities, how to accurately identify people’s travel mode has become particularly important for urban traffic planning and management. However, traditional methods are based on telephone interviews or questionnaires, which makes it difficult to obtain accurate and effective data. Nowadays, numbers of smartphones are equipped with various sensors, including accelerometers, gyroscopes, and GPS, providing a novel social sensing data source to detect people’s travel modes. The fusion of multiple sensor data is a promising way for travel mode detection. However, how to use these sensor data to accurately detect travel mode is still a challenging task. In this paper, we presented a light-weight method for travel mode detection based on four types of smartphone sensor data collected from an accelerometer, gyroscope, magnetometer, and barometer, and a prototype application was developed. Then, a novel convolutional neural network (CNN) was designed to identify five representative travel modes (walk, bicycle, bus, car, and metro). We compared the overall performance of the proposed method via different hyperparameters, and the experimental results show that the F value of the proposed method reaches 97%, which verified the effectiveness of the proposed method for travel mode classification.
ISSN:2076-3417
2076-3417
DOI:10.3390/app12136511