Employing Savitzky-Golay Smoothing in a Low Cost eHealth Platform
Ioannis E. Kosmadakis, Nikos Petrellis, Michael Birbas, M. Vardakas · 2018
The Savitzky-Golay (SG) filtering is evaluated in a low cost e-health platform that has been recently presented by the authors. The sensors' precision is not guaranteed in such a low cost platform since these sensors are not medically certified. The accuracy of the sensor measurements has to be improved in order to have a reliable e-health platform. Filtering methods like moving average window (MAW), principal component analysis (PCA) and a simplified Kalman filter have already been used for smoothing and for sensor precision improvement. The comparison of these methods with SG filtering shows that the latter can offer a significant precision improvement without the problems offered by other methods like the latency of the MAW. More specifically, the SG filtering can offer an up to 10 times better Normalized Mean Square Error (NMSE) or 25dB higher Signal to Noise Ratio (SNR) than the original values. PCA on the other hand offers a remarkable precision improvement in many cases (e.g., half NMSE or 5dB better SNR) along with 20% compression that leads to lower power consumption.