Abstract

Detecting P300 signal plays a vital role in Brain-Computer Interface (BCI) systems. In recent years, deep neural networks have been vastly utilized for P300 detection. However, by increasing the number of dimensions, the ratio of saddle points to local minima increases exponentially which hampers the performance of these networks for P300 detection. In this paper, a new method is introduced which improves the training procedure of Deep Belief Networks (DBN) by using Nesterov momentum. In another version of the proposed algorithm, the drop-based learning rate is also utilized to improve its performance. These strategies lead to escape saddle points in large scales of dimensions which causes the considerable improvement in P300 detection. Performance of the proposed algorithm is evaluated on a dataset of EPFL BCI group and compared with some other state-of-the-art methods. The results demonstrate that the best recognition rate is obtained by using the proposed method which is equal to 93.47%. The obtained results also show approximately 4.83% superiority in the accuracy of proposed algorithm against its alternatives.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call