Performance evaluation and comparison of filters for real time embedded system applications

This paper focuses on an evaluation of an exponentially weighted moving average (EWMA) filter which has its results compared to the well known Kalman filter (KF) and a proportional (P) filter. The KF is well known and widely used in a variety of applications because of its accuracy, robustness and relative computational simplicity. Scenarios however do exist where the KF is not a suitable filter to use due to perhaps computational resource restrictions or smooth results requirements where the KF's performance can be lacking. The EWMA filter is proposed as a viable alternative to the KF, by providing a similar accuracy, improved robustness and dynamic response whilst being substantially less computationally expensive. A P filter is also presented as an effective and robust filter with minimal computational requirements. Performance comparisons are made between the three filter types and specific use cases are highlighted for that will enable designers to select the most appropriate filter for their needs based on signal-to-noise ratios.