The main subject of authors' research are non-contact methods of glass breaks detection based on analysis of acoustic signal generated during phenomena. Problem has essential meaning for modern, cost effective alarm systems, particularly installed into big buildings. Signal has stochastic character and the main difficulty of the problem is variability of many parameters (e.g. size and thickness of glass pane, distance from window to detector) and big amount of false signals (mainly accidental glass hits without break). Authors developed detection algorithm which uses Wavelet Transformation and few selected measures for signal features extraction and classification. Obtained detection efficiency >90percent is satisfactory, but resistance to false signals (near to 80percent) does not fulfill assumed level. Because the Wavelet Packet Decomposition (WPD) allows more detailed analysis in frequency domain than WT, it is more suitable for extraction of time-frequency interdependencies in analyzed signals. This paper discuss some methods and results of WPD application and wavelet selection for improving system performance, and increasing the resistance to false signals.
展开▼
