A new dynamically convergent differential neural network for brain signal recognition. (January 2022)

• Record Type: Journal Article
• Authors: Zhang, Zhijun; Sun, Jiansheng; Chen, Tao
• Is Part Of: Biomedical signal processing and control. Volume 71(2022)Part A
• Publication Date: 2022-01
• Subjects: Neural dynamic -- Neural network -- Brain-computer interface -- Convergence -- Machine learning; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2021.103130
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: An ensemble dynamically convergent differential neural network (E-DCDNN) is proposed for P300 brain signal classification, which can be applied to brain-computer interface (BCI) systems. Being different from most of existing neural networks with gradient descent method, the proposed E-DCDNN speeds up the training process and achieves higher accuracy. Being different from the traditional data preprocessing with downsampling method which loses useful original signals, a moving average method on time window is proposed to make sure that all the signals work in this work. What is more, principal components analysis (PCA) method is used for feature selection and dimensionality reduction. To overcome the effect of classifier variability fully, in preprocessing step, 45 classifiers are constructed based on all different combinations of all training data. Then integrated average of all classifiers is used for classification. Comparative experiments on Data Set IIb of BCI Competition II and Data Set II of BCI Competition III with state-of-art methods, i.e., ESVM, EFLD, CNN-I, MCNN-I, BN3, ECNN and MsCNN, verify that the proposed E-DCDNN achieves 100 Abstract: In this paper, a dynamically convergent differential neural network (DCDNN) is proposed for P300 brain signal classification, which can be applied to brain-computer interface (BCI) systems. Being different from most of state-of-the-art neural networks with thousands of parameters to adjust, the proposed neural network possesses fast training rate since it is a lightweight neural network. In the framework, data preprocessing includes time window interception, bandpass filtering, moving average filtering and training set balance. After that, feature extraction based on PCA is used, and an ensemble DCDNN (E-DCDNN) is proposed for classification. The proposed framework can effectively increase signal-to-noise ratio (SNR), reduce classification dimension without distortion, and eliminate classifier variability. In this work, the proposed framework achieves 100% and 98% accuracy on Data Set IIb of BCI Competition II and Data Set II of BCI Competition III, which mean the state-of-the-art character recognition performance in existing methods.