A.Yu. Aksenov - Junior Research, SPIIRAS

In the modern approach to the audio-visual data transmission the accent is shifting towards the use of programmable digital multi-interface technologies aimed to work in the infocommunication systems. The main feature of the multimedia content is that the consumer estimates the quality not primarily on technical parameters, but by the presence of semantic content and its relevance to expectations, and on the extent of its impact on the mind. This creates the preconditions for the emergence of the phenomenon of information redundancy. Information redundancy in the multimedia data on the one hand is a prerequisite for the data compression but on the other hand opens the possibilities for different types of information security threats. Transmission and distribution of audio data, particularly in virtual form, comprising the step of compression needed to reduce the bit volume. Modern coders-decoders (codecs) with the lossy compression are using a psychoacoustic model that provides a lossy audio compression without significant degradation of sound quality due to audio redundancy elimination. The psycho-acoustic model considered in this paper for audio data redundancy estimation is a model used by Lame audio codec which is a modified version of the model originally developed by the Fraunhofer Institute (standard psychoacoustic model). The idea of the proposed method for estimating of the audio data information redundancy data is based on inverted condition of informativeness used in a standard psycho-acoustic model. In this case, it is possible to form a characteristic audio file consisting of only those components that the original audio codec is "recognized" uninformative. Using the obtained characteristic file allows estimation of information redundancy, spatially localize artifacts contained in the test audio fragment, produce non-referential quality assessment of audio codecs. The use of information redundancy features can improve the performance of data compression methods used in the processing of multimedia data, and to identify and reduce the risk of fraudulent actions made over digital content.