Parallel audio features extraction for sound indexing and retrieval systems

In this paper, experimental results of a proposed software library for audio features extraction is presented. This library is capable of indexing and retrieving a relatively wide class of audio signal thanks to the computing of a large number of descriptors. Taking into account the data dependences between audio descriptors, a number of parallel regions has been identified. Applying of OpenMP standard in the library core and executing the code parts concurrently improved the overall computation performance. The conducted experiments show that the proposed approach shorten the computational time up to above 60 percent in comparison with its sequential counterpart. This allowed us to use this library in practical audio analysis tasks.