Output details
11 - Computer Science and Informatics
University of Ulster
A time–frequency convolutional neural network for the offline classification of steady-state visual evoked potential responses
<24>This work was carried out under the Marie Curie European Transfer of Knowledge grant BrainRobot, to increase the reliability of BCI commands. The new convolutional neural network architecture presented has been adapted subsequently for the detection of multi-oriented shapes (work presented at the International Conference of Document Analysis and Recognition 2013). In particular, the approach has been tested for the recognition of handwritten digits in Bangla (an Indian script). The approach is still an active research topic for the creation of features that are invariant to geometric deformations.