Identification of Fatigue Drivers Based on Multiple Convolutional Neural Networks in Accelerometry Data

Driving ability may be negatively impacted by prolonged sitting and lack of sleep. Therefore, knowing a driver's objective sitting and sleeping patterns might assist to lessen potential dangers. Study participants' raw accelerometry information was collected throughout a simulated driving activity, and deep learning was used to categories participants' sitting and sleeping habits. The students work in the lab for a whole week. During a 20-minute simulated drive, raw accelerometry data was acquired via a device worn on the thigh. Accelerometry data was trained on two convolutional neural networks to create four distinct categories. Five-fold cross-validation was used to assess accuracy. Using class activation mapping, researchers were able to identify class-specific differences in the dynamics of movement and posture. Results from a simulated drive using a thigh-mounted accelerometer show that CNN isa viable option for categorization. The results of this method might help in the detection of potentially impaired drivers due to exhaustion.