The need to increase the signal-to-noise ratio is a significant task in almost all sensor applications that are exposed to a significant influence of ambient noise. The most common methods used to reduce ambient and/or measurement noise are based on the use of analog and digital filters or wavelet signal decomposition. The paper describes a new approach to signal filtering based on the application of the so-called a statistical filter that was realized by applying statistics of higher order, i.e. the fourth order cumulants. The statistical filter is successfully tested in analysis various types of sensor signals such as: acoustic, hydro acoustic, seismic, and other.