A fast feature extraction system on compressed audio data

An efficient system is described, which directly extracts features from compressed audio material. It consists of a time/frequency conversion method and a feature extraction algorithm. The conversion method provides the feature extraction algorithm with a suitable complex spectral representation directly from the compressed domain. It further allows to trade-off between computational complexity and conversion accuracy. Several operating points using different conversion accuracies were tested with an MPEG audio identification system in order to evaluate the identification confidence. Based on these results it is possible to reduce the computational complexity from O(N log N) to O(N) compared to the conventional approach (complete decoding followed by a frequency analysis).