Embedded hardware-based adaptive filtering for noise reduction in bioimpedance data.

With the rise of wearable, affordable solutions using integrated circuits like the AD5933, noise reduction in bioimpedance data has become increasingly important. In this paper, we present an automated method for the realization of a digital filter for noise reduction in bioimpedance data. Unlike traditional methods that require manual tuning, our approach automatically adjusts the filter coefficients based on the characteristics of the incoming bioimpedance data - specifically by minimizing the smoothness difference between consecutive filtered data points. This allows for optimal noise reduction without prior knowledge of the signal's characteristics. Using both synthetic and experimentally obtained impedance data, we demonstrated an up to 8 dB improvement in signal-to-noise ratio with noise levels of up to 2 %. The method was successfully implemented on a microcontroller board, with power consumption below 11 mW (@3.3 V) during filter operation and an execution time under 185 ms (@ 64 MHz).These results highlight the method's potential for wearable and portable applications. The versatility of the proposed method to different biological signals was demonstrated by successfully filtering electromyography (EMG) and respiration bioimpedance signals from human volunteers.