Parametric modelling of decomposed subbands: Resolution improvement and applications for narrow-band signals

The introduction of a zoom technique which combines the advantages of subband decomposition and linear prediction for spectral-analysis applications is discussed. The efficiency of the technique is tested under different transform lengths, different linear prediction orders, different frequency components, and different SNR. As the number of subbands increases, the spectral resolution improves accordingly without increasing the transform length. The technique shows an equivalent or even superior performance, with the advantage of its reduced complexity.