Advanced method of Nystagmus-phase separation using adaptive modification of time-frequency signal representation

The present paper introduces an advanced method of adaptative signal filtration in the time-frequency domain using reversible wavelet transformations for the precise delimitation of the nystagmus quick and slow phases in an electronystagmogram. In widely used automated ENG systems, the imprecision in phases delimitation-caused by constant-characteristics signal filtration-is the main source of inaccuracies in diagnostics parameters. The adaptive modification of filtration characteristics allows the highest frequency components to be cut-out, only when necessary, that means according to the local signal properties. The use of the reversible wavelet transform of the ENG signal makes it possible to modify each “atom” of signal independently. A significant improvement in phase separation has been achieved both for model and real signals