Lossless Audio Coding Using Code-Excited Linear Prediction with Embedded Entropy Coding

A hybrid lossy/lossless audio coding is presented in this paper. The lossy coding part is based on code excitation approach where the excitation gain is adapted in a sample-by-sample fashion. Two short term predictors are used to cope with rapid time-varying nature of audio signals. The error signal between the input and the synthetic signal is sent to the decoder after arithmetic coding to achieve lossless compression. The excitation codebook is searched by an M-L tree search strategy with joint minimum error energy and minimum code length as search criteria. The advantage of the proposed coder is its low decoding complexity due to simple code excitation structure, and the compression performance achieved is comparable to other advanced lossless coders.