TY - GEN
T1 - EEG Sparse Representation Based Alertness States Identification Using Gini Index
AU - Tageldin, Muna
AU - Al-Mashaikki, Talal
AU - Bali, Hamza
AU - Mesbah, Mostefa
PY - 2018
Y1 - 2018
N2 - Poor alertness experienced by individuals may lead to serious accidents that impact on people’s health and safety. To prevent such accidents, an efficient automatic alertness states identification is required. Sparse representation-based classification has recently gained a lot of popularity. A classifier from this class typically comprises three stages: dictionary learning, sparse coding and class assignment. Gini index, a recently proposed method, was shown to possess a number of properties that make it a better sparsity measure than the widely used l0- and l1-norms. This paper investigates whether these properties also lead to a better classifier. The proposed classifier, unlike the existing sparsity-based ones, embeds the Gini index in all stages of the classification process. To assess its performance, the new classifier was used to automatically identify three alertness levels, namely awake, drowsy, and sleep using EEG signal. The obtained results show that the new classifier outperforms those based on l0- and l1-norms.
AB - Poor alertness experienced by individuals may lead to serious accidents that impact on people’s health and safety. To prevent such accidents, an efficient automatic alertness states identification is required. Sparse representation-based classification has recently gained a lot of popularity. A classifier from this class typically comprises three stages: dictionary learning, sparse coding and class assignment. Gini index, a recently proposed method, was shown to possess a number of properties that make it a better sparsity measure than the widely used l0- and l1-norms. This paper investigates whether these properties also lead to a better classifier. The proposed classifier, unlike the existing sparsity-based ones, embeds the Gini index in all stages of the classification process. To assess its performance, the new classifier was used to automatically identify three alertness levels, namely awake, drowsy, and sleep using EEG signal. The obtained results show that the new classifier outperforms those based on l0- and l1-norms.
KW - Alertness classification
KW - Gini index
KW - Sparse representation
UR - http://www.scopus.com/inward/record.url?scp=85058989680&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85058989680&partnerID=8YFLogxK
U2 - 10.1007/978-3-030-04239-4_43
DO - 10.1007/978-3-030-04239-4_43
M3 - Conference contribution
AN - SCOPUS:85058989680
SN - 9783030042387
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 478
EP - 488
BT - Neural Information Processing - 25th International Conference, ICONIP 2018, Proceedings
A2 - Ozawa, Seiichi
A2 - Leung, Andrew Chi Sing
A2 - Cheng, Long
PB - Springer-Verlag
T2 - 25th International Conference on Neural Information Processing, ICONIP 2018
Y2 - 13 December 2018 through 16 December 2018
ER -