Multirate spectral analysis for passive sonar

Because of widespread application in audio and image coding, subband structures are now well understood. Multirate subband implementations provide significant computational advantages while reducing distortion caused by sample rate changes. Some so-called perfect reconstruction systems can even eliminate all aliasing distortion caused by tandeming multirate spectral analysis and synthesis, as is the case in coding applications. Despite the widespread popularity in coding, the potential of multirate subband structures for general purpose spectral analysis seems to be largely overlooked. This paper reviews one such structure-the multirate filter bank-and demonstrates its utility as a spectrum analyzer of signals, with particular emphasis on passive sonar. The main advantage of the multirate filter bank is that it provides adjustable time-frequency resolution, and in fact, is shown to be a generalization of the Discrete Fourier Transform. Issues of aliasing, computation, and latency are discussed and passive sonar data sets are used for illustration.