다중 출력 예측을 적용한 EEG 기반 Valence-Arousal 회귀 모델
Valence-Arousal Regression Model Using EEG with Multi-Output Prediction
0
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
Unlike previous classification-based approaches that discretize continuous emotional states, this study pro- poses a multi-output regression model for predicting emotional valence and arousal using electroencephalography (EEG) signals. This method captures the continuous and complex nature of emotions. We compared three approaches: single-output regression, which predicts valence and arousal separately, and multi-output regression and multi-output regression using chain structure, both of which predict valence and arousal simultaneously. The single-output approach, though simple and easy to analyze, showed limitations in handling complex emotional states and required two model parameters, leading to inefficiencies. However, the multi-output approach effectively handled both emo- tional dimensions simultaneously, offering a more efficient and streamlined solution while maintaining high accuracy. Both LGGNet and CCNN models demonstrated the ability to accurately predict emotional states in the valence- arousal space, closely matching actual emotional states. This indicates that the regression models can effectively cap- ture complex emotions. Future work will explore more advanced network architectures to further improve emotion recognition accuracy.
Abstract
Unlike previous classification-based approaches that discretize continuous emotional states, this study pro- poses a multi-output regression model for predicting emotional valence and arousal using electroencephalography (EEG) signals. This method captures the continuous and complex nature of emotions. We compared three approaches: single-output regression, which predicts valence and arousal separately, and multi-output regression and multi-output regression using chain structure, both of which predict valence and arousal simultaneously. The single-output approach, though simple and easy to analyze, showed limitations in handling complex emotional states and required two model parameters, leading to inefficiencies. However, the multi-output approach effectively handled both emo- tional dimensions simultaneously, offering a more efficient and streamlined solution while maintaining high accuracy. Both LGGNet and CCNN models demonstrated the ability to accurately predict emotional states in the valence- arousal space, closely matching actual emotional states. This indicates that the regression models can effectively cap- ture complex emotions. Future work will explore more advanced network architectures to further improve emotion recognition accuracy.
참고문헌 (23)
- Hasnul MA, "Electrocardiogram-based emotion recognition systems and their applications in healthcare—a review", Sensors, 21(15), pp.015~ (2021)
- Nakasone A, "Emotion recognition from electromyography and skin conductance", Conf Proc Biosignal Interpretation, pp.219~222 (2005)
- Robinson S, "FMRI techniques and protocols", pp.411~456 (2009), City
- Lan Z, "Using support vector regression to estimate valence level from EEG", Conf Proc IEEE Int Conf Syst Man Cybern, pp.002558~002563 (2016)
- Torres EP, "EEGbased BCI emotion recognition : A survey", Sensors, 20(18), pp.5083~ (2020)
- Ishino K, "A feeling estimation system using a simple electroencephalograph", Conf Proc IEEE Int Conf Syst Man Cybern, 5, pp.4204~4209 (2003)
- Russell JA, "A circumplex model of affect", J Pers Soc Psychol, 39(6), pp.1161~ (1980)
- Tao W, "EEGbased emotion recognition via channel-wise attention and self attention", IEEE Trans Affect Comput (2020)
- Lin YP, "EEG-based emotion recognition in music listening: A comparison of schemes for multiclass support vector machine", Conf Proc IEEE Int Conf Acoust Speech Signal Process, pp.489~492 (2009)
- Soleymani M, "Continuous emotion detection in response to music videos", Conf Proc IEEE Int Conf Autom Face Gesture Recognit, pp.803~808 (2011)
- Garg N, "Decoding the neural signatures of valence and arousal from portable EEG headset", Front Hum Neurosci, 16, pp.1051463~ (2022)
- Galvão F, "Predicting exact valence and arousal values from EEG", Sensors, 21(10), pp.3414~ (2021)
- Torres-Valencia CA, "Multiple-output support vector machine regression with feature selection for arousal", valence space emotion assessment, 2014, 970~973
- Wang Y, "A multi-label learning method for efficient affective detection", Conf Proc IEEE EMBS Int Conf Biomed Health Inform, pp.61~64 (2017)
- Wu D, "Affect estimation in 3D space using multi-task active learning for regression", IEEE Trans Affect Comput, 13(1), pp.16~27 (2019)
- Wang Y, "A multi-label learning method for efficient affective detection", Conf Proc IEEE EMBS Int Conf Biomed Health Inform, pp.61~64 (2017)
- Kuppens P, "The relation between valence and arousal in subjective experience varies with personality and culture", J Pers, 85(4), pp.530~542 (2017)
- Morris JD, "SAM : The self-assessment manikin; an efficient cross-cultural measurement of emotional response", J Advert Res, 35(6), pp.63~68 (1995)
- Koelstra S, "DEAP : A database for emotion analysis using physiological signals", IEEE Trans Affect Comput, 3(1), pp.18~31 (2012)
- Alakus TB, "Database for an emotion recognition system based on EEG signals and various computer games – GAMEEMO", Biomed Signal Process Control, 60, pp.101951~ (2020)
- Yang Y, "Continuous convolutional neural network with 3D input for EEG-based emotion recognition", Conf Proc Int Conf Neural Inf Process, pp.433~443 (2018)
- Ding Y, "LGGNet : Learning from local-global-graph representations for brain–computer interface", IEEE Trans Neural Netw Learn Syst (2023)
- Alvarez, M, "Sparse convolved Gaussian processes for multi-output regression", Adv Neural Inf Process Syst, 21 (2008)