Available online at www.ijournalse.org Emerging Science Journal (ISSN: 2610-9182) Vol. 7, No. 1, February, 2023 Page | 116 Continuous Capsule Network Method for Improving Electroencephalogram-Based Emotion Recognition I Made Agus Wirawan 1, 2 , Retantyo Wardoyo 1* , Danang Lelono 1 , Sri Kusrohmaniah 3 1 Department of Computer Science and Electronics, Faculty of Mathematics and Natural Science, Universitas Gadjah Mada, Yogyakarta 55281, Indonesia. 2 Education of Informatics Engineering Department, Faculty of Engineering and Vocational, Universitas Pendidikan Ganesha, Singaraja 81116, Indonesia. 3 Department of Psychology, Faculty of Psychology, Universitas Gadjah Mada, Yogyakarta 55281, Indonesia. Abstract The convolution process in the Capsule Network method can result in a loss of spatial data from the Electroencephalogram signal, despite its ability to characterize spatial information from Electroencephalogram signals. Therefore, this study applied the Continuous Capsule Network method to overcome problems associated with emotion recognition based on Electroencephalogram signals using the optimal architecture of the (1) 1 st , 2 nd , 3 rd , and 4 th Continuous Convolution layers with values of 64, 128, 256, and 64, respectively, and (2) kernel sizes of 2×2×4, 2×2×64, and 2×2×128 for the 1 st , 2 nd , and 3 rd Continuous Convolution layers, and 1×1×256 for the 4 th . Several methods were also used to support the Continuous Capsule Network process, such as the Differential Entropy and 3D Cube methods for the feature extraction and representation processes. These methods were chosen based on their ability to characterize spatial and low-frequency information from Electroencephalogram signals. By testing the DEAP dataset, these proposed methods achieved accuracies of 91.35, 93.67, and 92.82% for the four categories of emotions, two categories of arousal, and valence, respectively. Furthermore, on the DREAMER dataset, these proposed methods achieved accuracies of 94.23, 96.66, and 96.05% for the four categories of emotions, the two categories of arousal, and valence, respectively. Finally, on the AMIGOS dataset, these proposed methods achieved accuracies of 96.20, 97.96, and 97.32% for the four categories of emotions, the two categories of arousal, and valence, respectively. Keywords: Electroencephalogram; Emotion Recognition; Differential Entropy; Baseline Reduction; 3D Cube; Capsule Network; Continuous Convolution. Article History: Received: 09 June 2022 Revised: 17 August 2022 Accepted: 14 September 2022 Available online: 07 November 2022 1- Introduction Emotions are psychological reactions to daily social interactions [1], which emerge due to certain conditions or problems encountered in achieving the desired target [2]. It is categorized into arousal and valence, with positive and negative values, respectively. Valence is an individual's reaction toward an event, while arousal is the excitement to behave accordingly or express the feeling [3]. A combination of arousal and valence labels is categorized into four quadrants: the 1 st represents high arousal and positive valence (HAPV), 2 nd depicts high arousal and negative valence (HANV), 3 rd implies low arousal and negative valence (LANV), while the 4 th describes low arousal and positive valence (LAPV) [4, 5]. The emotional reactions in each quadrant represent mental health and human performance [6]. It is essential to recognize the emotions in each quadrant to understand these individuals' mental state and performance. * CONTACT: rw@ugm.ac.id DOI: http://dx.doi.org/10.28991/ESJ-2023-07-01-09 © 2023 by the authors. Licensee ESJ, Italy. This is an open access article under the terms and conditions of the Creative Commons Attribution (CC-BY) license (https://creativecommons.org/licenses/by/4.0/).