An entropy fusion method for feature extraction of EEG
Feature extraction is a vital part in EEG classification. Among the various feature extraction methods, entropy reflects the complexity of the signal. Different entropies reflect the characteristics of the signal from different views. In this paper, we propose a feature extraction method using the fusion of different entropies. The fusion can be a more complete expression of the characteristic of EEG. Four entropies, namely a measure for amplitude based on Shannon entropy, a measure for phase synchronization based on Shannon entropy, wavelet entropy and sample entropy, are firstly extracted from the collected EEG signals. Support vector machine and principal component analysis are then used for classification and dimensionality reduction, respectively. We employ BCI competition 2003 dataset III to evaluate the method. The experimental results show that our method based on four entropies fusion can achieve better classification performance, and the accuracy approximately reaches 88.36 %. Finally, it comes to the conclusion that our method has achieved good performance for feature extraction in EEG classification.
KeywordsEEG Feature extraction Entropy Feature fusion
The work is supported by National Natural Science Foundation of China (Nos. 61172134, 61671197 and 61601162).
Compliance with ethical standards
Conflict of interest
The authors declare that they have no conflict of interest.
- 5.Gaur P, Pachori RB, Wang H et al (2015) An empirical mode decomposition based filtering method for classification of motor-imagery EEG signals for enhancing brain-computer interface. In: Proceeding of 2015 international joint conference on neural networks (IJCNN) IEEE, pp 1–7Google Scholar
- 6.Marchesotti S, Bassolino M, Serino A et al (2016) Quantifying the role of motor imagery in brain-machine interfaces. Scientific reports, p 6Google Scholar
- 11.Xiao D, Mu Z, Hu J (2009) Classification of motor imagery EEG signals based on energy entropy. In: Proceeding of 2009 international symposium on intelligent ubiquitous computing and education, pp 61–64Google Scholar
- 12.Rui Z, Peng X, Rui C et al (2015) Predicting inter-session performance of SMR-based brain-computer interface using the spectral entropy of resting-state EEG. Brain Topogr 28(5):1–11Google Scholar
- 15.Zhang Z, Zhou Y, Chen Z et al (2013) Approximate entropy and support vector machines for electroencephalogram signal classification. Neural Regen Res 8(20):1844Google Scholar
- 16.Chen Z, Zhou H, Zhao L (2011) Decoding human right and left hand motor imagery from EEG single trials using sample entropy. In: Proceeding of 2011 international conference on IEEE electronics and optoelectronics (ICEOE), pp V4-353–V4-356Google Scholar
- 24.Tuncay C (2010) Entropy analyses of spatiotemporal synchronizations in brain signals from patients with focal epilepsies. arXiv preprint. arXiv:1002.3552
- 25.Bashar SK, Bhuiyan MIH (2015) Automatic feature selection based motor imagery movements detection scheme from EEG signals in the dual tree complex wavelet transform domain. In: Proceeding of 2015 IEEE international conference on telecommunications and photonics (ICTP), pp 1–5Google Scholar
- 26.Li X, Cui W, Li C (2012) Research on classification method of wavelet entropy and fuzzy neural networks for motor imagery EEG. In: Proceeding of 2012 IEEE international conference on modelling, identification & control (ICMIC), pp 478–482Google Scholar
- 32.Wang L, Xu G, Yang S et al (2012) Motor imagery BCI research based on sample entropy and SVM. In: Proceeding of 2012 sixth international conference on electromagnetic field problems and applications (ICEF), pp 1–4Google Scholar
- 33.Imran SM, Talukdar MTF, Sakib SK et al (2014) Motor imagery EEG signal classification scheme based on wavelet domain statistical features. In: Proceeding of 2014 international conference on electrical engineering and information communication technology (ICEEICT), pp 1–4Google Scholar
- 34.Gupta SS, Soman S, Raj PG et al (2014) Improved classification of motor imagery datasets for BCI by using approximate entropy and WOSF features. In: Proceeding of 2014 international conference on signal processing and integrated networks (SPIN), pp 90–94Google Scholar