The Stamp method for processing of high noise data from infrared sensor in harsh environment

•A new digital method, called Stamp method, for elimination of undesired noise data from a periodic signal is presented.•It is a combination of equidistant transversal and differential filters with subtraction of neighboring statistical images.•Influence of imperfection of data processing is checked by simple shifting of resulting data vector.•The method was applied on signal from IR sensor in harsh environment of hydroelectric power plant generator.

This article presents a new digital method, called Stamp method, for elimination of undesired noise data from a periodic IR signal. The Stamp method is a useful tool for statistical analysis of changes which appear in a noised input signal. One stamp is a statistical image of several consequence periods from a raw input signal. The Stamp method is a combination of equidistant transversal and differential filters, where useful output data is produced by subtraction of neighboring stamps. The corresponding digital Stamp filter formulation is provided with the Z-transform formalism. An application of the statistical Stamp method is presented on temperature changes extraction from a real noised IR (infra-red) raw signal from turning shaft in a power plant. The moving average and impulse-response filters were applied for final cleaning up of the data. The MATLAB short source code of stamp implementation is in the Appendix A.