QSVM-QNN: Quantum Support Vector Machine Based Quantum Neural Network Learning Algorithm for Brain-Computer Interfacing Systems

Bikash K. Behera, Saif Al-Kuwari, Ahmed Farouk · 2025

Read QSVM-QNN: Quantum Support Vector Machine Based Quantum Neural Network Learning Algorithm for Brain-Computer Interfacing Systems by Bikash K. Behera, Saif Al-Kuwari, Ahmed Farouk (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2505-14192.txt.

Introduction

Bikash Behera, Saif Al-Kuwari, and Ahmed Farouk join two quantum learners for brain-signal reading in a 2025 paper. Brain-computer links let minds steer kit with no hands, which could aid the sick and open fresh ways to work machines. Their reads stay shaky: waves vary, sorting runs slow, and each user's tune shifts in real time.

Their blend mates a quantum support method, sharp at drawing split lines, with a quantum web that learns rich shapes. The pair sorts brain-wave tasks with more grace than each half alone. Sturdier sorts could make brain links day-to-day tools.

Catalog record

CREATORBikash K. Behera, Saif Al-Kuwari, Ahmed Farouk
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE48 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2505-14192

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

QSVM-QNN: Quantum Support Vector Machine Based Quantum Neural Network Learning Algorithm for Brain-Computer Interfacing Systems Bikash K. Behera, Saif Al-Kuwari, Ahmed Farouk Abstract A brain-computer interface (BCI) system enables direct communication between the brain and external devices, offering significant potential for assistive technologies and advanced human-computer interaction. Despite progress, BCI systems face persistent challenges, including signal variability, classification inefficiency, and difficulty adapting to individual users in real time. In this study, we propose a novel hybrid quantum learning model, termed QSVM-QNN, which integrates a Quantum Support Vector Machine (QSVM) with a Quantum Neural Network (QNN), to improve classification accuracy and robustness in EEG-based BCI tasks. Unlike existing models, QSVM-QNN combines the decision boundary capabilities of QSVM with the expressive learning power of QNN, leading to superior generalization performance. The proposed model is evaluated on two benchmark EEG datasets, achieving high accuracies of 0.990 and 0.950, outperforming both classical and standalone quantum models. To demonstrate real-world viability, we further validated the robustness of QNN, QSVM, and QSVM-QNN against six realistic quantum noise models, including bit flip and phase damping. These experiments reveal that QSVM-QNN maintains stable performance under noisy conditions, establishing its applicability for deployment in practical, noisy quantum environments. Beyond BCI, the proposed hybrid quantum architecture is generalizable to other biomedical and time-series classification tasks, offering a scalable and noise-resilient solution for next-generation neurotechnological systems. † † impactstatement: Integrating quantum computing and machine learning techniques into brain-computer interface (BCI) systems represents a transformative step towards overcoming existing challenges in signal processing and classification accuracy. By harnessing the capabilities of quantum parallelism and entanglement, these innovative approaches have the potential to significantly enhance the efficiency and accuracy of BCI systems, ultimately enabling more seamless interaction between individuals and external devices. The promising results obtained from applying the proposed quantum algorithm, which integrates the quantum neural network (QNN) and the quantum support vector machine (QSVM), to EEG-based BCI datasets demonstrate their superiority over classical methods in terms of accuracy and robustness. I Introduction A new frontier in human-machine interaction is opened up by brain-computer interface (BCI) systems, which establish a direct connection between the brain and external devices. Electroencephalography (EEG) has emerged as a prominent non-invasive method to obtain brain signals and is at the core of BCI systems [ 1 ] . These systems interpret and analyze brain activity and translate it into valuable commands using sophisticated signal processing techniques and decoding algorithms [ 2 ] . The role of the user interface is essential to allow interaction and provide customized sensory interfaces, auditory input, and visual displays. Various types of BCI systems exist, such as P300-based BCIs, SSVEP BCIs, invasive BCIs, and hybrid BCIs [ 3 ] . These systems have different applications in areas such as healthcare, neurorehabilitation, neuroprosthetics, assistive technology, gaming, communication devices, and cognitive enhancement [ 4 ] . BCI systems face numerous obstacles, necessitating the incorporation of machine learning (ML) methods to improve their performance and dependability. One crucial challenge is the poor signal-to-noise ratio (SNR) in EEG and other neural recording technologies, which makes it difficult to correctly extract meaningful brain activity patterns [ 5 ] . Furthermore, inter- and intra-subject variability in brain signals requires adaptive learning models that generalize between users and conditions [ 6 ] . Real-time decoding is additionally challenged by the non-stationary nature of brain signals, which alter over time due to fatigue, attention shifts, or environmental influences [ 7 ] . Furthermore, the high dimensionality of brain data requires feature extraction and dimensionality reduction techniques to increase computational efficiency and classification accuracy [ 8 ] . These issues have resulted in the widespread use of ML techniques such as deep learning (DL) and reinforcement learning (RL). These techniques can improve signal processing and feature extraction and enable more resilient and adaptive BCI systems [ 9 ] . ML techniques have enabled BCI devices to overcome difficulties by dramatically improving accuracy and efficiency. Traditional classifiers have been frequently used for EEG-based BCI applications due to their simplicity and effectiveness in handling small datasets [ 10 ] . However, these models struggle with non-linear and multidimensional brain signals. More advanced techniques have shown better performance in automatically extracting significant features from raw EEG signals and capturing temporal dependencies [ 11 ] . Long short-term memory (LSTM) networks improve real-time decoding by solving the vanishing gradient problem in sequential neural data [ 12 ] . Furthermore, transfer learning approaches have been used to reduce inter-subject variability, allowing models trained on a single group of users to generalize to other individuals [ 13 ] . RL has also been investigated for adaptive BCI systems, which allow real-time tweaking of control signals based on user feedback [ 14 ] . Incorporation of deep learning (DL) and hybrid ML models continues to improve BCI performance, making it more robust and reliable for real-world applications like neuroprosthetics and assistive technology. Recent studies have demonstrated the effectiveness of deep learning approaches, particularly CNN-based models, for EEG signal classification. For example, the work in [ 15 ] proposed a multiscale CNN architecture to enhance spatial spectral feature extraction, achieving high accuracy in motor imagery tasks. Another study [ 16 ] employed a deep fusion of CNN and GRU for cross-subject emotion recognition, addressing generalization issues in EEG-based systems. Furthermore, in [ 17 ] , a hybrid attention-based CNN-LSTM model was introduced, showing significant improvements in interpretability and classification performance on EEG emotion datasets. Despite the immense advances achieved by ML in BCI systems, various major obstacles remain, necessitating the investigation of quantum machine learning (QML) as a feasible solution. One key challenge is the large dimensionality and complexity of brain signal data, which makes feature selection and classification computationally intensive for typical machine learning methods [ 18 ] . Furthermore, ML-based BCI models have inadequate subject generalization due to inter-individual heterogeneity in brain activity, resulting in uneven performance between users [ 19 ] . Furthermore, traditional DL algorithms require massive labeled datasets, making them data-hungry and computationally costly during training and adaptation. These limitations have driven the potential of QML to process high-dimensional neural data more efficiently than classical methods, leading to increased robustness and reduced computational complexity in BCI applications [ 20 ] . By utilizing QML, BCI systems could improve their feature extraction, classification, and real-time signal processing performance, making them more adaptive and dependable for neurotechnology applications. I-A Motivation QML has emerged as a feasible approach to solving the problems of classical ML in BCI systems. Several QML algorithms, including Quantum Support Vector Machines (QSVMs) and Quantum Neural Networks (QNNs), have shown potential to overcome the problem of categorization and feature extraction [ 21 ] . QSVMs use quantum kernel approaches to efficiently process high-dimensional data, whereas QNNs use quantum parallelism to improve pattern recognition and decision-making processes [ 22 ] . Despite these advances, current QML models frequently lack adaptability and robustness when dealing with non-stationary and noisy inputs, resulting in inferior performance in real-world applications [ 23 ] . To address this research gap, we propose a hybrid quantum-classical model combining QSVMs and QNNs to improve classification accuracy and adaptability in BCI systems. The QSVM component identifies complicated non-linear patterns from EEG data

Read the complete text