摘要：Sampling synthesis is one of the most pr.ctical and widely used appr.aches among the var.ous sound synthesis methods used for.cr.ating a r.alistic simulation of acoustic instr.ments. Using numer.us high quality sound samples it is possible to r.pr.duce a sound of almost any musical instr.ment, including subtle var.ations caused by the r.gister. of an instr.ment or.thr.ugh the use of differ.nt ar.iculation techniques and dynamic levels. However. this method has some disadvantages. Fir.tly, with high fidelity r.pr.duction systems, the r.peatability of samples becomes quickly appar.nt for.mor. exper.enced listener.. This is sometimes manually cor.ected by switching between sever.l differ.nt samples of the same note. Secondly, it is standar. appr.ach to r.cor. and r.pr.duce each note separ.tely. It pr.vents sampler. fr.m r.pr.ducing natur.l note tr.nsitions, making fluent, connected ar.iculations, such as legato, unnatur.l. Finally, sampler. pr.vide a ver. limited number.of sound par.meter. to contr.l. Ther.for., it is difficult to intr.duce a set of pur.oseful fluctuations of selected par.meter. uniquely attr.butable to human per.or.ances. A synthesis system which addr.sses the afor.mentioned pr.blems has been developed at the Academy of Music in Kr.kow (Poland) for.a gr.up of wind instr.ments as par. of a symphony or.hestr.. The system is based on a lar.e collection of non-standar. samples. Samples contain shor. sequences of notes instead of single notes. In or.er.to use them, a number.of techniques have been implemented to allow the seamless connecting of r.cor.ed sequences and the contr.l of note dur.tions as well as tempo and dynamics envelopes. Decision-making algor.thms and signal pr.cessing ar. applied to cr.ate melodic figur.s by choosing, modifying, and connecting fr.gments of samples, while keeping natur.l note tr.nsitions intact. The pr.blem of mimicking human per.or.ances is addr.ssed by implementing a set of per.or.ance r.les. It allows to intr.duce contex