Lossless compression of high fidelity audio and adaptive linear prediction

We consider the lossless compression of high fidelity (e.g. 16-bit) digital audio using adaptive linear prediction. Both linear predictive coding (LPC) and least mean squares (LMS) predictors are considered. Preliminary results are presented for the compression of industry standard Sound Quality Assessment Material (SQAM) samples from 16 bits to 1.5-3 bits. Previous results by others on the same audio source was in the 8-bit range.