We present a novel concept of "noncontiguous" audio segmentation by exploiting the high-level music structure. The existing lossless audio compressors working in asymmetrical mode divide the audio into quasi-stationary segments of variable length by recursive splitting (MPEG-4 ALS) or by dynamic programming (asymmetrical OptimFROG) before computing a set of linear prediction coefficients for each segment. Instead, we combine several variable length segments into a group and use a single set of linear prediction coefficients for each group. The optimal algorithm for combining has exponential complexity and we propose a quadratic time approximation algorithm. Integrated into asymmetrical OptimFROG, the proposed algorithm obtains up to 1.20% (on average 0.23%) compression improvements with no increase in decoder complexity.
Click to purchase paper as a non-member or login as an AES member. If your company or school subscribes to the E-Library then switch to the institutional version. If you are not an AES member and would like to subscribe to the E-Library then Join the AES!
This paper costs $33 for non-members and is free for AES members and E-Library subscribers.