Development Of A Mobile-Based Accident Detection And Notification System With Multi Modal Data
No Thumbnail Available
Date
2024
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Corpus Intellectual
Abstract
Road accidents is one of the factors responsible for fatalities
and lifelong disabilities globally. The predominant increase in
mortality rate on the highway has been attributed to the
late arrival of emergency authorities, which could be
because of delays in reporting time. The literature has
suggested a number of approaches to deal with this
significant issue. Some of these approaches include the use
of machine learning algorithms. However, many of these
works have been limited to single-mode (video, audio, or
image) accident detection methods. In addition, some of
the existing study explore the use of intelligence
transportation system, which could be considered expensive
to use. This research has developed a mobile-based
accident detection and notification system with multi-modal
data. The detection system has incorporated trained
simulated data that is validated by unseen data online. The
simulated data was characterized by some selected modal
data (accelerometer, sound, and gravitational force
magnitude) equivalent to the online features. The simulated
data was trained using the multilayered perceptron artificial
neural network (ANN) model. The trained model was tested
using online data from the data world platform. The
observation with the simulated data showed that the model
achieved an accuracy of 99.5%. The result of the experiment
on the online data showed 99.8% accuracy. At the end of
the modelling, the ANN model was integrated into an
android mobile-based accident detection and notification
system. Furthermore, the system was tested with several case
study and the result showed that the system performed as
expected.
Keywords: accelerometer, accident, artificial neural
network, audio, G-force, android mobile application.