[en] A characteristic feature of acoustic leakage detection systems is that they operate under conditions of high-level industrial noise, which cannot be eliminated by simple means at the measurement stage. The development of algorithms which would yield estimates stable with regard to noise constitutes a pressing problem. There exists another feature of acoustic systems for leakage detection based on statistical methods -- the necessity of processing prolonged implementations of random processes to obtain a satisfactory statistical accuracy. It would be of interest to devise estimates involving small error over a small implementation length, which would reduce the time of leakage detection. The aim of this work was to develop methods of obtaining noiseproof, rapid estimates. Discussions are included on providing an estimate insensitive to additive noise and on obtaining estimates insensitive to multiplicative noise. 2 refs., 4 figs