Time-frequency analysis applied to door slam sound quality problems

Abstract
Spectral analysis of non-stationary phenomena using traditional, segment averaged, discrete Fourier transform techniques suffers from limitations related to the trade-off between temporal and spectral resolution. These limitations are inherently due to the assumption of signal stationarity over an observation interval of fixed duration. In case frequency components occur in widely different frequency bands, an alternative technique, based upon the Wavelet transform, yields results with a temporal resolution which is inversely proportional to the analysis frequency. The issue of non-stationary spectral analysis is discussed at the hand of a sound quality problem related to a door slam. It is shown how the Wavelet analysis allows to identify the critical time instances and hence the critical physical phenomena responsible for the problem.