Traitement du signal · 2019 · 36 citations · 45 references
ComputingBiomedical EngineeringComputer ScienceComputational NeuroscienceMachine LearningNeural EngineeringBrain-computer InterfacesDeep LearningElectroencephalographyNeural InterfacesNeural NetworksSupport Vector MachinesBrain-computer Interface ClassificationBrain-computer Interface Signal ProcessingBrain-machine InterfacesClassification Algorithms For Brain-computer InterfacesConvolutional Neural NetworksDeep Learning For ElectroencephalogramSupport Vector ClassificationSupport Vector Machine Classification
The tongue is one of the few organs with high mobility in the case of severe spinal cord injuries. However, most tongue-machine interfaces (TMIs) require the patient to wear obtrusive and unhygienic devices in and around the mouth. This paper aims to develop a TMI based on the glossokinetic potentials (GKPs), i.e. the electrical signals generated by the tongue when it touches the buccal walls. Ten participants were recruited for this research. The GKP patterns were classified by convolutional neural network (CNN) and support vector machine (SVM). It was observed that the CNN outperformed the SVM in individual and average scores for both raw and preprocessed datasets, reaching an accuracy of 97~99%. The CNN-based GKP processing method makes it easy to build a natural, appealing and robust TMI for the paralyzed. Being the first attempt to process GKPs with the CNN, our research offers an alternative to the traditional brain-computer interfaces (BCIs), which suffers from the instability and low signalto-noise ratio (SNR) of electroencephalography (EEG).
45
EmotionMeter: A Multimodal Framework for Recognizing Human Emotions
Wei‐Long Zheng, Wei Liu, Bao‐Liang Lu et al. · IEEE Transactions on Cybernetics · 2018 · 1.2K citations