Biomedical Signal Processing and Control 44 (2018) 75–81 Contents lists available at ScienceDirect Biomedical Signal Processing and Control journal homepage: www.elsevier.com/locate/bspc Fuzzy adaptive neurofeedback training: An efficient neurofeedback training procedure providing a more accurate progress rate for trainee Nasrin Shourie a,∗ , Mohammad Firoozabadi b , Kambiz Badie c a Department of Biomedical Engineering, Science and Research Branch, Islamic Azad University, Tehran, Iran b Faculty of Medical Sciences, Tarbiat Modares University, Tehran, Iran c Research Institute for ICT, Tehran, Iran a r t i c l e i n f o Article history: Received 11 December 2016 Received in revised form 9 December 2017 Accepted 18 February 2018 Available online 22 April 2018 Keywords: Neurofeedback Threshold Scoring index Fuzzy technique Mental fatigue a b s t r a c t In this paper, a new fuzzy adaptive neurofeedback training procedure (FNFT) is proposed, in which a more effective performance in neurofeedback training can be expected. In the proposed FNFT, the threshold is adaptively set considering the cortical activity of the subject. Scoring index (SI) (the number of points increased in subject’s score) is set according to the brain activity of the subject and is calculated using a fuzzy rule based system. When training feature surpasses the threshold, the SI points are then added to the points of the subject. This adaptive scoring index leads to having an efficient indicator for the success rate of the subject. In addition, the subject is rewarded with an audio or visual feedback. The sound intensity of the audio feedback and the length and width of the video frame are adjusted in accordance with the SI. Finally, an EEG feature is also considered (brain mental fatigue index) to stop the training as the subject becomes mentally fatigued. © 2018 Published by Elsevier Ltd. 1. Introduction Neurofeedback is a non-invasive conditioning method, in which individuals can learn to voluntarily regulate their brain activities [1–3]. Athletes, artists and business executives take advantage of neurofeedback to learn how to use the full potential of their brain to reach their superior performance. Neurofeedback is also effective enough to treat the patients with anxiety, depression, epilepsy, etc [3–7]. Neurofeedback training generally consists of recording EEG sig- nals from one or two electrode sites and providing audio or visual feedback for the individuals about their cortical activities [8,9]. Training features are calculated on a moving window that is contin- uously updated and compared with a threshold. The subject would be rewarded with audio or visual feedback or point increasing when the training feature surpasses the threshold. It must be noted that inappropriate scoring method may confuse subjects when evaluat- ing how successful they were in modifying their brain activity. Researchers have previously applied various rewarding meth- ods for NFT. One of the most traditional thresholding methods is to have a threshold fixed by a therapist. In this method, the therapist ∗ Corresponding author. E-mail addresses: shourie.n@srbiau.ac.ir (N. Shourie), pourmir@modares.ac.ir (M. Firoozabadi), k badie@itrc.ac.ir (K. Badie). manually sets a fixed threshold for a training session. If the training feature surpasses the threshold, a point would then be added to the score of the subject [10,11]. An open question is how the threshold should be selected. Some therapists set the threshold according to the previous results of the subject and their sensitivity to rewards and punishments [10]. Other therapists determine the threshold according to a schedule upon which the threshold is set to a mini- mum value in the first session and then is increased during the next sessions. However, a fixed threshold setting would not be adaptive regarding the brain activity of subjects. Therefore, the thresholds should be adaptively modified according to the cortical activity of the subject. In addition, the number of points and the audio or visual feedback should accurately reflect the progress rate of the subject. Another traditional thresholding method is automatic calcula- tion, done by setting the threshold to the training feature level surpassing 60–85% of the time during the preceding 30 s moving average window [1,8,9,12–18]. When the training feature surpasses the determined threshold, the score would increase by one point and the subject would then be rewarded with an audio or visual feedback. This method is adaptive in nature, and there is no need for the threshold to be manually set by the therapist. However, this method may confuse the subject and therefore, not provide an appropriate indicator of the success rate. Suppose that it is aimed to train a subject to increase his/her alpha wave activity. For this purpose, the subject starts with a default thresh- old and then passes the first 30 s window. It is assumed that the new https://doi.org/10.1016/j.bspc.2018.02.009 1746-8094/© 2018 Published by Elsevier Ltd.