Related Experiment Videos
Convolutional Neural Network-Based Classification of Driver's Emotion during Aggressive and Smooth Driving Using
Kwan Woo Lee1, Hyo Sik Yoon2, Jong Min Song3
1Division of Electronics and Electrical Engineering, Dongguk University, 30 Pildong-ro 1-gil, Jung-gu, Seoul 100-715, Korea. leekwanwoo@dgu.edu.
Sensors (Basel, Switzerland)
|March 24, 2018
Summary
Detecting aggressive driving early is crucial for safety. This study introduces a new method using facial images and artificial intelligence to accurately identify aggressive driver emotions, outperforming existing techniques.
Area of Science:
- Computer Science
- Artificial Intelligence
- Human-Computer Interaction
Background:
- Aggressive driving causes significant societal harm, necessitating proactive detection of adverse driver emotional states.
- Existing methods using accelerometers, gyroscopes, electroencephalography (EEG), or electrocardiogram (ECG) sensors have limitations, including driver discomfort, signal detachment, GPS dependency, and susceptibility to environmental interference.
Purpose of the Study:
- To develop a novel method for detecting aggressive driving by analyzing driver emotions.
- To overcome the limitations of current driver emotion detection systems.
Main Methods:
- A convolutional neural network (CNN)-based approach was employed to analyze driver emotions.
- Facial images captured using near-infrared (NIR) light and thermal cameras served as input data.
- A custom database was utilized for training and evaluating the model.
Main Results:
- The proposed CNN-based method achieved high classification accuracy in detecting driver emotions indicative of aggressive or smooth driving.
- The system demonstrated superior performance compared to existing driver emotion detection methodologies.
Conclusions:
- Facial emotion detection using NIR and thermal imaging with CNNs offers a promising, non-invasive, and robust solution for identifying aggressive driving.
- This approach effectively addresses the shortcomings of previous sensor-based and GPS-dependent systems.