Signal windowing is a temporal weighting operation whereby a signal is multiplied by a function that pays more emphasis on desired parts of the signal and typically attenuates it outside this span, normally to zero, in order to result in a finite support (nonzero part). In audio and acoustics applications it is quite common that better frequency resolution is desirable at low frequencies, while at high frequencies better time resolution calls for a shorter analysis or processing window. In this paper methods to realize and utilize frequency-dependent signal windowing are presented that exhibit this desirable frequency-dependent property. Particularly methods based on time-frequency warping and rewarping of signals are discussed. Audio examples are presented where this technique is found useful.
