Perceptually Optimized Cascaded Long Term Prediction of Polyphonic Signals for Enhanced MPEG-AAC

Nanjundaswamy T, Rose K. Perceptually Optimized Cascaded Long Term Prediction of Polyphonic Signals for Enhanced MPEG-AAC. In: AES Convention 131. Audio Engineering Society; 2011. Paper 8518. Available from: https://aes.org/publications/elibrary-page/?id=16044

@inproceedings{Nanjundaswamy2011_16044,
  author = {Nanjundaswamy, Tejaswi and Rose, Kenneth},
  title = {{Perceptually Optimized Cascaded Long Term Prediction of Polyphonic Signals for Enhanced MPEG-AAC}},
  booktitle = {AES Convention 131},
  note = {Paper 8518},
  year = {2011},
  month = oct,
  publisher = {Audio Engineering Society},
  url = {https://aes.org/publications/elibrary-page/?id=16044}
}

TY  - CPAPER
TI  - Perceptually Optimized Cascaded Long Term Prediction of Polyphonic Signals for Enhanced MPEG-AAC
AU  - Nanjundaswamy, Tejaswi
AU  - Rose, Kenneth
T2  - AES Convention 131
M1  - Paper 8518
PY  - 2011
DA  - 2011/10/06
UR  - https://aes.org/publications/elibrary-page/?id=16044
PB  - Audio Engineering Society
LA  - en
AB  - MPEG-4 Advanced Audio Coding uses the long term prediction (LTP) tool to exploit inter-frame correlations by providing a segment of previously reconstructed samples as prediction for the current frame, which is naturally useful for encoding signals with a single periodic component. However, most audio signals are polyphonic in nature containing a mixture of several periodic components. While such polyphonic signals are themselves periodic with overall period equaling the least common multiple of the individual component periods, the signal rarely remains sufficiently stationary over the extended period, rendering the LTP tool ineffective. Further hindering the LTP tool is the typically employed parameter selection based on minimizing the mean squared error as opposed to the perceptual distortion criteria defined for audio coding. We thus propose a technique to exploit the correlation of each periodic component with its immediate past, while taking into account the perceptual distortion criteria. Specifically, we propose cascading LTP filters corresponding to individual periodic components, designed appropriately in a two stage method, wherein an initial set of parameters is estimated backward adaptively to minimize the mean squared prediction error, followed by a refinement stage where parameters are adjusted to minimize the perceptual distortion. Objective and subjective results validate the effectiveness of the proposal on a variety of polyphonic signals.
ER  -