Modified locally-adaptive myriad filters

Abstract

Modified myriad locally-adaptive filters (LAF) in a sliding data window with complex use of local activity indicators (LAIs) of previously designed myriad LAFs are proposed. Improvement of integral indicator for the whole complex one-dimensional signal and local efficiency indicators at segments of linear change, constant level and step edge of signal function was achieved according to criterion of the mean-square error. Based on LAIs comparison, the myriad LAFs switch adaptive myriad filters with adaptation of linearity parameter K depending upon local scale estimates and with different values of window length and coefficient influencing on K. In one of the proposed LAFs, the LAI are also used for approximate estimation of the noise level and for choosing the corresponding set of component filters with parameters more appropriate to the estimated noise level. The preliminary use of the robust Hampel myriad filter is proposed for better spikes removal. The algorithms are implemented in real time.