Realization of current-mirror based wavelet filters for biomedical signal processing
Loading...
Date
2026-02-11
Authors
Σεννής, Αναστάσιος
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Electroencephalogram (EEG) signal processing plays a critical role in the detection and analysis of neurological disorders, such as epilepsy, where signals are often weak, noisy, and highly non-stationary. Traditional digital processing techniques, although powerful, may suffer from high power consumption and latency, making them less suitable for wearable and implantable biomedical devices. This Thesis presents the design, implementation, and experimental validation of a low-power analog wavelet filter for real-time EEG signal analysis, with emphasis on epileptic event detection. The work begins with the theoretical development of the Marr wavelet filter, whose transfer function is analyzed and decomposed using partial fraction expansion. MATLAB simulations are used to validate the frequency and time-domain behavior of the proposed filter and to establish the design parameters for hardware implementation. The wavelet filter is then realized using current-mode circuits based on CMOS current mirrors, enabling low-voltage operation, high bandwidth efficiency, and reduced power consumption. First-order low-pass filters and integrators are combined to form second order sections, which are subsequently cascaded to construct a fourth-order wavelet filter. Extensive circuit-level simulations demonstrate the filter’s ability to accurately detect epileptic events in real EEG signals. Following the schematic design, the filter is implemented at the layout level using TSMC65nmCMOStechnology. Post-layout simulations confirm that the filter maintains its functionality despite parasitic effects, while achieving compact area and robust performance. Finally, the proposed wavelet filter is implemented using Field Programmable Analog Array (FPAA) technology to validate the design experimentally. Measured results from real hardware demonstrate successful detection of epileptic events in raw EEG signals, highlighting the flexibility and effectiveness of reconfigurable analog signal processing.The results of this work demonstrate that analog wavelet filtering using current mirrors and FPAA technology is a viable and efficient approach for low-power biomedical signal processing, paving the way for future real-time, adaptive, and wearable EEG monitoring systems.
Description
Keywords
EEG signal processing, Programmable wavelet filters, Marr wavelet filter, Current-mirrors, Cadence simulations, Low-power design, Biomedical signal processing, Monte Carlo analysis, Field programmable analog array (FPAA)