Conception of hierarchical dynamic structure in application to audio and video data recognition

摘要

The wide range of signal processing methods such as wavelets and fractal decomposition or texture analysis usually misses the vital components - the semantic structure. In our approach to the structuring of audio and video data, which was realized in a software package Semantic Analysis of Images (SAI), we apply the adaptive dynamic data structure for object-fitting hierarchical analysis of audio and video data. This package presents the new software tools to reveal the interrelated network of context independent semantics from the initial data structure. The theoretic basis of our approach is the localization of semantically important areas through the new rating principle and following iterative synthesis of hierarchical trees to create the coherent structure of selected fragments. Adaptive dynamic data structure, as the result of this process, contains the signal fragments, which are essentially important for the following processing. Most evident application of semantic decomposition is the preliminary structuration of audio and video data for subsequent application of object identification routines and target-oriented discriminating compression of images.