Adaptive_Filtering_for_EEG_Noise_Reduction
Adaptive filtering is a smart, real-time noise removal technique that continuously learns and removes unwanted signals from EEG.
Electroencephalogram (EEG) signals are of extremely low intensity, on the order of microvolts, and are highly vulnerable to various kinds of noise, including eye blinks (EOG), muscle activities (EMG), and power line noise. All of this severely affects the quality of the EEG signal. However, fixed filter banks do not efficiently filter out this noise from the EEG signal, as they are non-stationary and tend to overlap each other.
In this project, a new concept of designing a real-time adaptive filter for denoising the EEG signal using FPGA is to be implemented. It is based on using a specific adaptive filter algorithm, such as the Least Mean Squares (LMS) or Normalized LMS.
The architecture of the design includes an interface for acquiring the EEG signal, an analog-to-digital conversion block, and a digital processing unit implemented using FPGA technology. The digital processing unit includes an adaptive FIR filter core, an error computation unit, a coefficient update unit, and a control unit. The design also includes a reference noise signal, which in this case is an EOG signal, to enhance the performance of the noise cancellation process.
The design of the adaptive filter aims to provide an efficient way of canceling noise from the EEG signal, thus making the design efficient for portable devices used in real-time processing of EEG signals.
The performance of the design is evaluated based on parameters such as Signal-to-Noise Ratio (SNR) improvement, Mean Square Error (MSE), and hardware resource usage. The performance of the design has shown satisfactory results, thus validating the design for biomedical signal processing.