کنفرانس مهندسی برق ایران

صفحه اصلی / سی و سومین کنفرانس بین المللی مهندسی برق

Hand Movment Decoding from EEG Signals Using Kalman Filter with Parameters Estimated via Neural Networks and Least Squares Method

نویسندگان :

Pegah Khoshkavandi¹ Mohammad B Shamsollahi² Ali Motie Nasrabadi³

1- دانشگاه صنعتی شریف 2- دانشگاه صنعتی شریف 3- دانشگاه شاهد تهران

کلمات کلیدی :

Brain-computer interfaces،Kalman filter،Multilayer perceptron،Electroencephalography

چکیده :

Brain-computer interfaces (BCIs) facilitate direct communication between the brain and external devices, offering transformative potential for individuals with motor disabilities. One of the main challenges in this area is the accurate interpretation of hand movements from non-invasive electroencephalographic (EEG) signals, which are often affected by inherent noise and complexity. This study explores the integration of a Kalman filter with a multilayer perceptron (MLP) to enhance the estimation of hand movement trajectories based on EEG signals. While the Kalman filter is commonly used for continuous motion decoding, its effectiveness hinges on the precise analysis of its parameters, particularly the transfer matrix. Traditionally, these parameters are calculated using the least squares method. In this work, we propose a hybrid approach in which the transition matrix \mathbit{F}_\mathbit{i} is dynamically estimated using an MLP, while the remaining parameters are obtained via the least squares method. Using a 5-fold cross-validation protocol on EEG data from three individuals, the hybrid approach consistently showed improved correlation values for motion estimation across all axes when compared to the traditional Kalman filter. Notably, the Z-axis exhibited the most significant improvements, indicating that the hybrid approach effectively addresses the limitations of the Kalman filter. These findings highlight the potential of combining neural networks models with classical filtering techniques to achieve more accurate and reliable motion decoding. This advancement offers promising opportunities for brain-computer interfaces (BCIs) in assistive and rehabilitation technologies.