Pattern extraction in sparse representations with application to audio coding

This article deals with the extraction of frequency-domain auditory objects in sparse representations. To do so, we first generate sparse audio representations we call spikegrams, based on neural spikes using gammatone/gammachirp kernels and matching pursuit. We then propose a method to extract frequent auditory objects (patterns) in the afore-mentioned sparse representations. The extracted frequency-domain patterns help us address spikes (atoms or auditory events) collectively rather than individually. When audio compression is needed, the different patterns are stored in a small codebook that can be used to efficiently encode audio materials in a lossless way. The approach is applied to different audio signals and results are discussed and compared. Our experiments show that substantial coding gain is obtained when our technique based on pattern extraction is used as opposed to the case where spikes (atoms) are coded individually. This work is a first step towards the design of a high-quality “object-based” audio coder.