Using a learned algorithm for detecting a deterministic signal in a noise of unknown intensity

Abstract

The main sources of noise and interference, the features of detecting signals in noises, methods of dealing with them during transmission over various distances are considered. Passive methods of dealing with noise and interference are to reduce the length of cable networks to a reasonable minimum and reduce the number of cables; using only high quality cables and connectors from well-known manufacturers; laying cables with large bending radii to avoid interference from the so-called triboelectric effect; separation of signal and power cable trunks; using matched loads; in such a use of equipment so that its operating conditions are significantly lower than the limiting ones; in using the most interference-resistant interface. Active methods of dealing with noise and interference consist in the use of intermediate signal amplifiers that compensate for their attenuation in the line due to ohmic resistance and losses at high frequencies due to cable reactivity; switching to a twisted pair; switching to fiber-optic cable when it is necessary to transmit a signal over long distances. In a number of problems of signal reception in the presence of noise, one cannot be limited by such a general criterion as the signal-to-noise ratio. There is a need to use more subtle statistical properties of processes that allow one to quantify the reliability of the data. Noise is an important problem in science and technology because it defines lower limits both in terms of the accuracy of any measurement and in terms of the magnitude of signals that can be processed electronically. An adaptive detection of signals in noise is defined, using an optimal detection algorithm for analysis. An algorithm is proposed for detecting a signal in noise both in discrete observation and in the transition to continuous observation of signals. Probabilistic characteristics of signal detection in noise of unknown and known intensity of the learning algorithm are constructed.