Table 5 BCIs for communication functions
From: Recent applications of EEG-based brain-computer-interface in the medical field
Research field | Subject | Method | Effect evaluation | References |
|---|---|---|---|---|
Communication—speller—MI | 2 subjects | The first BCI speller system based on MI to enter text by changing the orientation of the device | Develop an effective synchronized BCI system that uses the minimum number of controls (2) to control 30 targets (26 letters + punctuation) | [135] |
3 subjects (Age: 24 – 26 years, males) | A BCI speller based on 4 control commands was designed using advanced EEG decoding and NLP techniques | Three subjects achieved spelling rates of 3, 2.7, and 2 characters/min, respectively | [136] | |
6 healthy subjects (Age:19 – 26 years, 4 males) | Design of an Oct-o-Spell paradigm incorporating intelligent input methods | PTE mode is more efficient than non-PTE mode | [137] | |
Communication—speller—P300 | 10 healthy male subjects | Designed a paradigm that includes modifications to the 9-key text interface | Significantly reduces word entry time and makes word entry easier | [138] |
10 participants [Age: (28 ± 4.84) years, 5 males] | A model is proposed that combines the two distinguishing features of 3D animation and the use of column flashes only | All participants declared in the subjective test that the proposed paradigm was more user-friendly than the classical paradigm | [139] | |
9 subjects (Age: 20 – 34 years, 4 males) | Designed the first truly gaze-independent visual BCI, the stimuli consisted of colors displayed in widescreen, which made gaze focus independent of BCI spelling | The speller was tested online. Using 5 repetitions of the stimulus, the mean online symbol selection accuracy was 88.4%, and the mean online spelling speed was 1.4 characters/min | [140] | |
25 subjects | A multi-window discriminative regular pattern matching method is proposed | The algorithm won first place in the 2019 World Brain-Controlled Robotics Competition | [141] | |
Dataset I: gigascience | ||||
Dataset II: scientific data | Designed filter banks and typical correlation analysis combination methods | Successful reduction of the number of P300 experiments good recognition rates and shorter recognition times in character recognition | – | [142] |
BNCI Horizon 2020 Database | – | |||
P300 Akimpech Database | Introducing a data alignment method in transfer learning | Data from different subjects can be more evenly distributed and inter-subject variability is effectively reduced | [143] | |
Communication—speller—SSVEP | 22 healthy subjects (Age: 23 – 28 years, 9 males) | Design of a 120-target BCI spelling system based on coded modulated visual evoked potentials | Has a high average message rate (265.74 bits/min), while the equipment is large, fast, and has a short training time | [145] |
10 subjects [Age: (30 ± 5) years] | Design of a spelling system incorporating EMG signals and SSVEP paradigms | Provide a user-friendly, practical system for speller applications | [146] | |
20 healthy participants (Age: 24 – 46 years, 16 males) | Proposed design of a stimulus-responsive hybrid speller using eye tracking with EEG and video | The authors indicated that this set produced less fatigue, worry, strain, and discomfort | [147] | |
12 healthy subjects (Age: 20 – 26 years, 5 males) | Improvements to the SSVEP typing system in three areas: the user interface, the EEG acquisition device, and the decoding algorithm | Make the typing system more flexible and more suitable for practical application scenarios | [148] | |
Communication—speller—hybrid | 14 healthy subjects (Age: 18 – 41 years, 8 males) | A synchronous hybrid system based on P300 and SSVEP | The average online utility information transfer rate achieved using this method was 53.06 bits/min | [149] |
11 healthy volunteers (Age: 19 – 24 years, males) | A hybrid speller in the frequency-enhanced row-column paradigm is proposed, which allows the simultaneous triggering of two signals: the P300 and the SSVEP | The average accuracy of the proposed hybrid BCI is 94.29% and the information transfer rate is 28.64 bits/min | [150] | |
20 subjects | A hybrid BCI-speller system is proposed, encoded by a combination of EEG functions: the P300, motor-visual evoked potentials, and SSVEP, and using a layout of more than 200 targets | The online prompted spelling and free spelling results show that the proposed hybrid BCI speller achieves average accuracies of (85.37 ± 7.49)% and (86.00 ± 5.98)% for the classification of 216 targets, with average message transfer rates of (302.83 ± 39.20) bits/min and (204.47 ± 37.56) bits/min, respectively | [144] | |
54 subjects | A hybrid BCI based on a two-stream convolutional neural network is presented | In single mode (70.2% for MI and 93.0% for SSVEP) and mixed mode (95.6% for MI-SSVEP mixed mode) | [151] | |
Communication-the handwriting paradigm | 5 healthy right-handed participants [Age: (30.83 ± 1.8) years, males] | Subjects were asked to write “Hello, world!” repeatedly on a tablet. Then machine-learning methods were used to recognize the EEG signals | The results suggest that the possibility of recognizing handwritten content directly from brain signals | [152] |
A participant with hand paralysis due to spinal cord injury | Demonstration of an intracortical BCI that attempts to decode handwritten actions from neural activity in the motor cortex and translate them into text in real time | The authors showed that the subjects’ typing speeds were comparable to the mobile phone typing speeds of smart people in the participants’ age group | [93] | |
The handwriting BCI dataset was publicly released by the Stanford University research team of Willett et al. [93] | A temporal channel cascade transformation network is proposed to decode neural activity for imagining handwriting movements | In the imaginary single character recognition task, the recognition accuracy of the proposed model can reach 95.78%, which is 2% better than the existing state-of-the-art model | [153] | |
Communication—the speech decoding | A speech-impaired individual with ALS | A chronically implanted BCI was used to synthesize comprehensible words online | Results show that 80% of synthetic words can be correctly recognized by human listeners | [154] |
An ALS patient | Implantation of electrocorticography into the sensorimotor cortex allows patients to operate computer applications with 6 intuitive voice commands | The results show that speech commands can be accurately detected and decoded without the need to recalibrate and retrain the model | [155] | |
48 participants (22 males) | Development of a novel differentiable speech synthesizer | The ability to process data with different spatial sampling densities and to simultaneously process EEG signals from the left and right brain hemispheres | [156] |