Method for reducing radar signal noise and computer

A method for filtering noise, intended to be implemented on a set of elementary amplitude signals each coming from a sensor (2), said sensor (2) being configured to:

• emit radiofrequency pulses, referred to as emitted pulses (Tx(k)), in the direction of a target (3), and periodically with a predetermined pulse emission period (Te), and;
• measure the amplitude of a received radiofrequency signal (Rx) resulting in particular from the reflection of the emitted radiofrequency pulses (Tx(k)) from said target (3), the received radiofrequency signal (Rx) consisting of received elementary signals (Rx(k)) each associated with a respective one of the emitted radiofrequency pulses (Tx(k)), the amplitude of each of the received elementary signals (Rx(k)) being measured for a plurality of sampling windows (Δt) and over a predetermined time range the origin of which is a time of emission of the corresponding emitted radiofrequency pulse, each sampling window (Δt) being associated with an index i that relates to a difference at the origin of the corresponding time range;
• the method being implemented by at least one computer (10), and the method comprising:
• a. said computer (10) receiving (S100) a plurality of elementary amplitude signals (S(k), 1<k<N1), each elementary amplitude signal being representative of the amplitude of a corresponding received elementary signal (Rx(k)),
• b. selecting a plurality of signals from among the elementary amplitude signals (S(k)) received in step a. so as to form a predetermined subset of elementary amplitude signals (S(k), k0<k<k0+N2) (S200),
• c. for each sampling window of index i, computing a variation indicator in relation to the elementary amplitude signals of said subset (S300),
• d. when each of a predetermined number of variation indicators is less than a predetermined threshold (Th) (S410), updating a noise vector (G), said updating comprising:
• d1. computing, for each sampling window of index i, an average M(i) for said subset of elementary amplitude signals (S(k, i), k0<k<k0+N2) (S310), and
• d2. constructing a new noise vector (G), where each component G(i) of the noise vector (G) is associated with one of the sampling windows of index i, and where each component G(i) of the noise vector (G) adopts a value that depends on the average M(i) computed in step d1 and associated with the same sampling window of index i (S420), and
• e. for at least one of the elementary amplitude signals (S(k)), computing a corrected elementary signal (Scorr(k1)) by subtracting the corresponding component G(i) of the noise vector from said elementary amplitude signal and for each sampling window of index i (S500).