Early distracted driving detection from smartphone audio signals

Abstract

In most of the road accidents cases registered, the prime cause always revolves around distracted driving. Distracted driving comprises of four types of driving events: Fetching Forward, Eating or Drinking, Turning Back, Picking Drops.This problem can be resolved by predicting the driver’s behavior and take preventive actions beforehand. To make this into reality, we have used concepts of deep learning. It use the recorded audio signals from the driver’s cellular devices, these signals are created during any of the above driving event occurs. It convert them into a frequency-time profile and then pass them to a classifier, which classifies them into four mentioned driving eventsbased on that preventive measure could be taken.