Research and analysis of performance evaluation metrics for estimation algorithms

Abstract

Filtering techniques has been widely used in digital signal processing field. For a great diversity of filtering algorithms, it is necessary to assess the performance of filters in different situations by using appropriate performance evaluation metrics. However, the adopted metrics are quite simplex generally in the measuring and evaluating of filtering algorithms, and the metrics mostly limit to RMSE (root mean square error) and ME (median error). To solve the problem, an objective and complete performance evaluation metrics system could be of great meaning. A performance evaluation metrics system based on Monte Carlo is established which synthesized metrics that are commonly used, and analysis and comparisons of emphasis points of each metrics are made in detail. The simulation result proved effectiveness of the proposed metrics system, in addition, demonstrated the advantage and disadvantage of these metrics in practical application.