On the performance similarity between exponential moving average and discrete linear kalman filter

Raw signal from sensor is generally corrupted by noise and other uncertainties. To suppress the noise, a filtering mechanism is required. Exponential Moving Average filter (EMA) serves as a powerful yet exceptionally simple filter. However, selecting so-called Alpha parameter in EMA is not a straightforward task. Extremely large Alpha value lead to more noise in the signal and small Alpha value resulted in sluggish convergence to the true value. In many cases Alpha parameter is selected arbitrarily and then opted for the best performance, resulting in numerous trials and errors. This paper is aimed to present an insight to select Alpha parameter, that is by implementing Kalman Gain from Kalman Filter structure as Alpha parameter and simultaneously highlight the similarity between Discrete Linear Kalman filter and Exponential Moving Average filter through some mathematical manipulations. To demonstrate the filter performance, altitude data from BMP280 barometric sensor is filtered. The results show that the EMA with Kalman Gain is capable to converge to the true altitude value and for some reasons, EMA with Kalman Gain resembles Kalman Filter performance in this particular scenario.