Alertness and mental fatigue classification using computational intelligence in an electrocardiography and electromyography system with off-body area network

• Main Authors: Concepcion, Ronnie S., Manalo, Jommel S., Garcia, Ave Jianne D., Legaspi, Rhaniel A., Prestousa, Jun Angelo, Pascual, Gio Paolo C., Firmalino, Junco S., Ilagan, Lorena C.
• Format: text
• Published: Animo Repository 2020
• Subjects: Mental fatigue; Electrocardiography; Electromyography; Computational intelligence; Electrical and Electronics; Systems and Communications
• Online Access: https://animorepository.dlsu.edu.ph/faculty_research/1549; https://animorepository.dlsu.edu.ph/context/faculty_research/article/2548/type/native/viewcontent
• Institution: De La Salle University

Preempting mental fatigue may cause decrease in the quality of life and the worst accidents. A system of electrocardiography and electromyography signals can enhance the detection of alertness and mental fatigue. This study determines the suitability of some computational intelligence, namely, artificial neural network (ANN), fuzzy logic system, and a Sugeno adaptive neuro-fuzzy inference system (ANFIS), in detecting mental alertness and fatigue of a person using neurophysiological signals of electrocardiogram (ECG) and electromyogram (EMG) only instead of using higher-dimensional array of physiological data. The usage of these neurophysiological signals was tested if it correlates with high detection rate as to the usual observable physiological parameters. Muscle contraction was also studied in parallel with varying heart rates. Moreover, a power-efficient off-body access network (oBAN) was materialized using Arduino microcontroller with Bluetooth wireless transmission medium. The system is composed of two major parts: the development of BAN and the implementation of soft algorithms. The data set was extracted from 20 university students of differing ages, genders, and sleep hours. Provided with the same training set, the system detection accuracy for ANN, FIS, and ANFIS is 97.800%, 99.529%, and 99.604%, respectively. An identical testing set was also employed to ANN, FIS, and ANFIS, yielding 71.000%, 99.553%, and 99.556% detection accuracy. Hence, with this physiological data set and purposive classification, ANFIS provides the paramount accuracy. © Springer Nature Switzerland AG 2020.