Skip to main content

Table 5 A comparison of the provided methods in other papers and the proposed method for Emotion Recognition

From: A novel approach to emotion recognition using local subset feature selection and modified Dempster-Shafer theory

Authors Year Method Classification accuracy (%)
Fan and Chou [66] 2018 Recurrence quantification analysis, logistic regression 75.7%
Zhong et al. [33] 2017 Spectral and time features, multiple-fusion-layer based ensemble classifier of stacked autoencoder (MESAE) 77.19% (arousal accuracy), 76.17% (valence accuracy)
Atkinson and Campos [22] 2016 Statistical and spectral features, Hjorth parameters, fractal dimension, minimum-Redundancy-Maximum-Relevance, support vector machine 62.39% (valence), 60.72% (arousal)
Xu and Plataniotis [32] 2016 Power spectral density, stacked denoising autoencoders, deep belief network 85.86% (arousal accuracy of SDAE), 84.77% (valence accuracy of SDAE), 88.33% (arousal accuracy of DBN), 88.59% (valence accuracy of DBN)
Jie et al. [67] 2014 Sample entropy, support vector machine 79.11%
Yin et al. [33] 2017 Spectral and time features, multiple-fusion-layer based ensemble classifier of stacked autoencoder 77.19% (arousal accuracy)
76.17% (valence accuracy)
Tripathi et al. [21] 2017 Convolutional neural networks, deep neural network 58.44% (valence, DNN), 55.70% (arousal, DNN), 66.79% (valence, CNN), 57.58% (arousal, CNN)
Alam et al. [29] 2016 Convolutional neural networks 81.17%
Kumar et al. [25] 2016 Bispectrum, least square support vector machine, radial basis function, linear neural network 64.86% (arousal), 61.17% (valence)
Our work 2018 The proposed method 90.54%