Search: onr:"swepub:oai:DiVA.org:mdh-53510" >
Driver’s Cognitive ...
Driver’s Cognitive Load Classification based on Eye Movement through Facial Image using Machine Learning
-
- Rahman, Hamidur, Doctoral Student, 1984- (author)
- Mälardalens högskola,Inbyggda system
-
- Ahmed, Mobyen Uddin, Dr, 1976- (author)
- Mälardalens högskola,Inbyggda system
-
- Barua, Shaibal (author)
- Mälardalens högskola,Inbyggda system
-
show more...
-
- Funk, Peter, 1957- (author)
- Mälardalens högskola,Inbyggda system
-
- Begum, Shahina, 1977- (author)
- Mälardalens högskola,Inbyggda system
-
show less...
-
(creator_code:org_t)
- English.
- Related links:
-
https://urn.kb.se/re...
Abstract
Subject headings
Close
- The driver's cognitive load is considered a good indication if the driver is alert or distracted but determing cognitive load is challenging and the acceptance of wire sensor solutions like EEG and ECG are not not preferred in real-world driving scenario. The recent development of image processing, machine learning, and decreasing hardware prices enables new solutions and there are several interesting features related to the driver’s eyes that are currently explored in research. Two different wireless sensor systems, one commercial giving eye position (SmartEye) and one Microsoft LifeCam Studio with resolution 1920 x 1080 were used for data collection. In this paper, two eye movement parameters, saccade, and fixation are investigated through facial images and 13 features are manually extracted. Five machine learning algorithms, support vector machine (SVM), logistic regression (LR), linear discriminant analysis (LDA), k-nearest neighbors (k-NN), and decision tree (DT), are investigated to classify the cognitive load. According to the results, the SVM model with linear kernel function outperforms the other four classification methods. Here, the achieved average accuracy is 92% using SVM. Again, three deep learning architectures, convolutional neural networks (CNN), long short-term memory (LSTM), and autoencoder (AE) are designed both for automatic feature extraction and cognitive load classification. The results show that CNN architecture achieves the highest classification accuracy which is 91%. Besides, the classification accuracy for the extracted eye movement parameters is compared with reference eye tracker signals. It is observed that the classification accuracies between the eye tracker and the camera are very similar.
Subject headings
- NATURVETENSKAP -- Data- och informationsvetenskap -- Datavetenskap (hsv//swe)
- NATURAL SCIENCES -- Computer and Information Sciences -- Computer Sciences (hsv//eng)
Keyword
- Computer Science
- datavetenskap
- Computer Science
- datavetenskap
Publication and Content Type
- vet (subject category)
- ovr (subject category)
To the university's database