Sequential forward mother wavelet selection method for mental workload assessment on N-back task using photoplethysmography signals

Aydemir T., ŞAHİN M., AYDEMİR Ö.

INFRARED PHYSICS & TECHNOLOGY, vol.119, 2021 (SCI-Expanded) identifier identifier

  • Publication Type: Article / Article
  • Volume: 119
  • Publication Date: 2021
  • Doi Number: 10.1016/j.infrared.2021.103966
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Academic Search Premier, Aerospace Database, Chimica, Communication Abstracts, Compendex, INSPEC
  • Keywords: Photoplethysmography, Mental workload, Wireless sensor, Classification, CLASSIFICATION, RECOGNITION
  • Karadeniz Technical University Affiliated: Yes


The increasing demands of a cognitive task require additional brain resources. This demand, known as mental workload, can lead to deteriorated task performance. Therefore, assessment of mental workload can provide a proper working environment to promote the working efficiency or improve safety in high-risk working environments for a subject. In this study, we present a novel sequential forward mother wavelet selection method for three levels of mental workload assessment on N-back task using photoplethysmography (PPG) signals, which non-invasively measures the blood volume changes in the microvascular bed of tissue from the skin surface with a low-cost opto-electronic technique. The proposed method was successfully applied to a PPG dataset, which was recorded from 22 healthy subjects during an N-back task using a wearable sensor. Instead of using only one mother wavelet, the features were extracted from effective mother wavelet combinations by means of a sequential forward mother wavelet selection method. In this three-class problem, the highest classification accuracy (CA) rates were achieved with 10 s (s) PPG signal segments compared with the 6 s, and 8 s PPG signal segments. For the 10 s PPG signals segments the highest CA was obtained as 76.67% for Subject 20 and the average CA for all subjects was obtained as 65.76%. Furthermore, the proposed method provided 3.59% CA improvement in average. We believed that the proposed method could ensure a great alternative to conventional mental workload assessment techniques.