Improving Electroencephalography-Based Imagined Speech Recognition with a Simultaneous Video Data Stream