LSTMNCP: lie detection from EEG signals with novel hybrid deep learning method

Aslan M., Baykara M., Alakuş T. B.

MULTIMEDIA TOOLS AND APPLICATIONS, vol.83, pp.31655-31671, 2024 (SCI-Expanded) identifier identifier

  • Publication Type: Article / Article
  • Volume: 83
  • Publication Date: 2024
  • Doi Number: 10.1007/s11042-023-16847-z
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus, FRANCIS, ABI/INFORM, Applied Science & Technology Source, Compendex, Computer & Applied Sciences, INSPEC, zbMATH
  • Page Numbers: pp.31655-31671
  • Karadeniz Technical University Affiliated: Yes


Lying has become an element of human nature. People lie intentionally or unintentionally at any point in their lives. Human beings can deceive by lying to justify themselves about something or to get rid of a wrongdoing. This lie can result in various consequences, including health deterioration, loss of life, a sense of insecurity, criminal behaviors, and more. Such situations are more common especially in daily life, security, and criminology. In these cases, lie detection is of vital importance. With the development of technology, lie detection becomes a more important issue. People can manipulate others and provide information by lying. This situation has led researchers to turn to more alternative ways and the importance of EEG signals has increased. Since EEG signals are difficult to manipulate, there has been an increase in their use and analysis in lie detection studies. In this study, lie detection was performed with EEG signals and the importance of EEG signals was demonstrated. Within the scope of this study, a novel hybrid deep learning method was designed on the Bag-of-Lies dataset, which was created using different methods, and lie detection was carried out. The study consisted of four stages. In the first stage, EEG data were obtained from the Bag-of-Lies dataset. In the second stage, the data were decomposed into sub-signals by DWT method. These signals, which were separated in the third stage, were classified with the designed novel hybrid deep learning model. At the last stage of the study, the performance of the classifier was determined by accuracy, precision, recall, F1-score, and AUC score. At the conclusion of the research, an accuracy score of 97.88% was achieved, demonstrating the significance of EEG signals in this domain.