This study harnesses EEG signals to enable the real-time control of servo motors, utilizing the OpenBCI Community Dataset to identify and assess brainwave patterns related to motor imagery tasks. Specifically, the dataset includes EEG data from 52 subjects, capturing electrical brain activity while participants imagined executing specific motor tasks. Each participant underwent multiple trials for each motor imagery task, ensuring a diverse and comprehensive dataset for model training and evaluation. A deep neural network model comprising convolutional and bidirectional long short-term memory (LSTM) layers was developed and trained using k-fold cross-validation, achieving a notable accuracy of 98%. The model’s performance was further compared against recurrent neural networks (RNNs), multilayer perceptrons (MLPs), and Τransformer algorithms, demonstrating that the CNN-LSTM model provided the best performance due to its effective capture of both spatial and temporal features. The model was deployed on a Python script interfacing with an Arduino board, enabling communication with two servo motors. The Python script predicts actions from preprocessed EEG data to control the servo motors in real-time. Real-time performance metrics, including classification reports and confusion matrices, demonstrate the seamless integration of the LSTM model with the Arduino board for precise and responsive control. An Arduino program was implemented to receive commands from the Python script via serial communication and control the servo motors, enabling accurate and responsive control based on EEG predictions. Overall, this study presents a comprehensive approach that combines machine learning, real-time implementation, and hardware interfacing to enable the precise and real-time control of servo motors using EEG signals, with potential applications in the human–robot interaction and assistive technology domains.

This research presents a machine learning model that predicts left, right, or no action using electroencephalography (EEG) signals extracted from two different wearable EEG headsets. The research aims to develop an accurate and efficient model by following a rigorous and effective process divided into two parts. In Part I, the constant features approach is employed, which involves data loading, feature extraction, preprocessing, model selection, and tuning the best model for optimal performance. The performance of classification algorithms (support vector machine (SVM), decision tree classifier, and random forest classifier) is evaluated using root-mean-squared error metrics. In Part II, the multivariate time series approach is utilized to improve the accuracy and robustness of the model. The approach involves data loading, preprocessing (such as normalizing the data), modeling, results analysis, and deployment preparation. A neural network architecture consisting of convolutional filters followed by a long short-term memory neural network (LSTM) is used in the proposed approach. The convolutional layer performs a convolution of an input series of feature maps with a filter matrix to extract high-level features. The LSTM network is specifically designed to capture long-term dependencies and overcome the issue of vanishing gradients. The proposed approach achieves an accuracy of 98% and can be used for real-time testing. The model can be utilized in various fields where accurate and real-time prediction of brain–computer interfaces (BCI) actions is crucial. Overall, the proposed approach provides a promising solution to the problem of action prediction using EEG signals, and further research can be conducted to explore its potential applications and optimize its performance.

The aim of this paper is to present the design and implementation of a brain-computer interface (BCI) system that can control a robotic claw. The system is based on the Emotiv Epoc headset, which provides the capability of simultaneous recording of 14 EEG channels, as well as wireless connectivity by means of the Bluetooth protocol. The system is initially trained to decode what user thinks to properly formatted data. The headset communicates with a personal computer, which runs a dedicated software application, implemented under the Processing integrated development environment. The application acquires the data from the headset and invokes suitable commands to an Arduino Uno board. The board decodes the received commands and produces corresponding signals to a servo motor that controls the position of the robotic claw. The system was tested successfully on a healthy, male subject, aged 28 years. The results are promising, taking into account that no specialized hardware was used. However, tests on a larger number of users is necessary in order to draw solid conclusions regarding the performance of the proposed system.

Brain-computer interfaces (BCIs) hold immense promise for human benefits, enabling communication between the brain and computer-controlled devices. Despite their potential, BCIs face significant cybersecurity risks, particularly from Bluetooth vulnerabilities. This study investigates Bluetooth vulnerabilities in BCIs, analysing potential risks and proposing mitigation measures. Various Bluetooth attacks such as Bluebugging, Bluejacking, Bluesnarfing, BlueBorne, Location Tracking, Man-in-the-Middle Attack, KNOB, BLESA and Reflection Attack are explored, along with their potential consequences on commercial BCI systems. Each attack is examined in terms of its modus operandi and effective mitigation strategies.

This study presents an in silico proof of concept exploring whether deep learning models can perform conditional mitochondrial DNA (mtDNA) sequence prediction across species boundaries. A CNN–BiLSTM model was trained under a leave-one-species-out (LOSO) scheme on complete mitochondrial genomes from 21 vertebrate species, primarily archosaurs. Model behavior was evaluated through multiple complementary tests. Under context-conditioned settings, the model performed next-nucleotide prediction using overlapping 200 bp windows to assemble contiguous 2000 bp fragments for held-out species; the resulting high token-level accuracy (>99%) under teacher forcing is reported as a diagnostic of conditional modeling capacity. To assess leakage-free performance, a two-flank masked-span imputation task was conducted as the primary evaluation, requiring free-running reconstruction of 500 bp interior spans using only distal flanking context; in this setting, the model consistently outperformed nearest-neighbor and demonstrated competitive performance relative to flank-copy baselines. Additional robustness analyses examined sensitivity to window placement, genomic region (coding versus D-loop), and random initialization. Biological plausibility was further assessed by comparing predicted fragments to reconstructed ancestral sequences and against composition-matched null models, where observed identities significantly exceeded null expectations. Using the National Center for Biotechnology Information (NCBI) BLAST web interface, BLASTn species identification was performed solely as a biological plausibility check, recovering the correct species as the top hit in all cases. Although limited by dataset size and the absence of ancient DNA damage modeling, these results demonstrate the feasibility of conditional mtDNA sequence prediction as an initial step toward more advanced generative and evolutionary modeling frameworks.