감정을 분류하는 대표적인 알고리즘에는 Support-vector-machine (SVM), Bayesian decision rule 등이 있다. 하지만 기존의 연구자들은 위와 같은 방법에는 문제점이 있다고 지적하였다. 이를 보완하기 위해 다른 연구자는 경사도를 이용하여 새로운 패턴인식 알고리즘을 제안하였다. 본 논문에서는 이 알고리즘을 통해 새로운 EEG 기반의 감정 인식 알고리즘을 제안하고 기존의 연구와 비교한다. 본 논문에서는 신뢰도 높은 자료를 얻기 위해 여러 논문에서 사용된 DEAP (a database for emotion analysis using physiological signals)를 사용하였다. 또한, 객관적인 검증을 위해 기존의 연구에서 사용된 4개의 뇌파 채널(Fz, Fp2, F3, F4)의 PSD (Power Spectral Density)를 특징으로 사용하여 감정의 2개 척도 (Arousal, Valence)를 분류하였다. 본 논문에서 실시한 교차검증 (4-fold)에 의하면 Valence 축에서 85%, Arousal 축에서 87.5의 정확도를 얻을 수 있었다.
감정을 분류하는 대표적인 알고리즘에는 Support-vector-machine (SVM), Bayesian decision rule 등이 있다. 하지만 기존의 연구자들은 위와 같은 방법에는 문제점이 있다고 지적하였다. 이를 보완하기 위해 다른 연구자는 경사도를 이용하여 새로운 패턴인식 알고리즘을 제안하였다. 본 논문에서는 이 알고리즘을 통해 새로운 EEG 기반의 감정 인식 알고리즘을 제안하고 기존의 연구와 비교한다. 본 논문에서는 신뢰도 높은 자료를 얻기 위해 여러 논문에서 사용된 DEAP (a database for emotion analysis using physiological signals)를 사용하였다. 또한, 객관적인 검증을 위해 기존의 연구에서 사용된 4개의 뇌파 채널(Fz, Fp2, F3, F4)의 PSD (Power Spectral Density)를 특징으로 사용하여 감정의 2개 척도 (Arousal, Valence)를 분류하였다. 본 논문에서 실시한 교차검증 (4-fold)에 의하면 Valence 축에서 85%, Arousal 축에서 87.5의 정확도를 얻을 수 있었다.
There are several algorithms to classify emotion, such as Support-vector-machine (SVM), Bayesian decision rule, etc. However, many researchers have insisted that these methods have minor problems. Therefore, in this paper, we propose a novel method for emotion recognition based on Electroencephalogr...
There are several algorithms to classify emotion, such as Support-vector-machine (SVM), Bayesian decision rule, etc. However, many researchers have insisted that these methods have minor problems. Therefore, in this paper, we propose a novel method for emotion recognition based on Electroencephalogram (EEG) signal using the Gradient method which was proposed by Han. We also utilize a database for emotion analysis using physiological signals (DEAP) to obtain objective data. And we acquire four channel brainwaves, including Fz (${\alpha}$), Fp2 (${\beta}$), F3 (${\alpha}$), F4 (${\alpha}$) which are selected in previous study. We use 4 features which are power spectral density (PSD) of the above channels. According to performance evaluation (4-fold cross validation), we could get 85% accuracy in valence axis and 87.5% in arousal. It is 5-7% higher than existing method's.
There are several algorithms to classify emotion, such as Support-vector-machine (SVM), Bayesian decision rule, etc. However, many researchers have insisted that these methods have minor problems. Therefore, in this paper, we propose a novel method for emotion recognition based on Electroencephalogram (EEG) signal using the Gradient method which was proposed by Han. We also utilize a database for emotion analysis using physiological signals (DEAP) to obtain objective data. And we acquire four channel brainwaves, including Fz (${\alpha}$), Fp2 (${\beta}$), F3 (${\alpha}$), F4 (${\alpha}$) which are selected in previous study. We use 4 features which are power spectral density (PSD) of the above channels. According to performance evaluation (4-fold cross validation), we could get 85% accuracy in valence axis and 87.5% in arousal. It is 5-7% higher than existing method's.
* AI 자동 식별 결과로 적합하지 않은 문장이 있을 수 있으니, 이용에 유의하시기 바랍니다.
가설 설정
Obtain the gradient vector; 3. Find the angle between a resultant vector and gradient; 4. Classify classes using the angle. It could complement the existing model (Naïve Gaussian classifier) and could obtain high accuracy result.
이론/모형
This algorithm that was proposed by Han[7]. The method assumes data distribution to be Gaussian (like parametric learning method) and it determines classes using the gradient of Gaussian functions. Another researcher investigated that this method could obtain higher accuracy than Naïve Bayesian (NB) method[7].
참고문헌 (14)
E. H. Han, and H. T. Cha., "A Study of Emotional Dimension for Mixed Feelings," Korea Journal of KOSES, Vol. 16, No. 4, pp. 469-480, 2013.
E. H. Han, and H. T. Cha., "A Study of Emotional Dimension that takes into account the Characteristics of the Arousal axis," Korea Journal of KOSES, Vol. 17, No. 3, pp. 57-64, 2014.
David. Lee, S. G. Lee, "The EEG Signal Based Emotion Recognition Using the EMD," Journal of KIIT, Vol. 12, No. 6, pp. 47-53, June 2014.
Blume. Warren. T, and Kaibara. Masako, Atlas of Pediatric, Philadelphia: Lippincott Williams & Wilkins.
Niedermeyer. E. and Lopes da Silva F., Electroencephalography: Basic Principles, Clinical Applications, and Related Fieldstlas of Pediatric, Philadelphia: Lippincott Williams & Wilkins.
I. S. Oh., Pattern Recognition, Seoul: Kyobobook Press.
E. H. Han, and H. T. Cha., "A Novel Method for Reliable Classifier using Gradient," IEEK Transactions on Smart Processing and Computing, Vol. 6, No. 2, 2016.
Christopher J. C. Burges, "A Tutorial on Support Vector Machines for Pattern Recognition," Data Min Knowl Discov, Vol. 2, No. 2, pp. 121-167, Jun. 1998.
E. H. Han, and H. T. Cha., "A Novel Method for Classifying Affective Terms using the Emotional Region Model," Korea Journal of KOSES, Vol. 20, No. 2, 2016.
S. Koelstra, et al., "DEAP: A Database for Emotion Analysis using Physiological Signals," IEEE Transaction on Affective Computing, Vol. 3, No. 4, pp. 18-31, 2012.
Bradley. M. M. and Lang. P.J, "Measuring emotion: the Self-Assessment Manikin and the Semantic Differential," Journal of Behav Ther Exp Psychiatry, Vol. 25, No. 1, pp. 49-59, Mar. 1994.
AI-Helper
※ AI-Helper는 부적절한 답변을 할 수 있습니다.