Bit Allocation for Linear Prediction Coefficients with Application to Lossless Audio Compression
We propose a novel technique of using bit allocation for linear prediction coefficients in asymmetric lossless audio compression. We show how to determine the optimal bit allocation using a new closed-form formula for the excess error from quantization and describe a recently introduced algorithm (Optimization-Quantization Least Squares) which computes the optimal quantized prediction coefficients applied for the allocation. The proposed method, implemented as a modified asymmetrical OptimFROG, obtains small (but consistent) signal dependent compression improvements with virtually no decoder complexity increase (for an 847 MB audio corpus, up to 0.27%, on average around 0.06%). Compared to MPEG-4 ALS, it obtained 0.38% better compression, while being at the same time approximately 5 times faster at decoding.
Click to purchase paper 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 $20 for non-members, $5 for AES members and is free for E-Library subscribers.