Rhythmic Music Codec: An MDCT-Based Music Codec with Onset-Localized Block Switching and Tempo-Synchronized Prediction

Bao H, Bukovec N, Krinsky S, Bosi M. Rhythmic Music Codec: An MDCT-Based Music Codec with Onset-Localized Block Switching and Tempo-Synchronized Prediction. In: AES Show 2026 Nashville. Audio Engineering Society; 2026.pp. 43:1-12. Available from: https://aes.org/publications/elibrary-page/?id=23474

@inproceedings{Bao2026_23474,
  author = {Bao, Haotian and Bukovec, Nick and Krinsky, Summer and Bosi, Marina},
  title = {{Rhythmic Music Codec: An MDCT-Based Music Codec with Onset-Localized Block Switching and Tempo-Synchronized Prediction}},
  booktitle = {AES Show 2026 Nashville},
  pages = {43:1-12},
  url = {https://aes.org/publications/elibrary-page/?id=23474},
  year = {2026},
  month = oct,
  publisher = {Audio Engineering Society},
  address = {Gaylord Opryland Resort}
}

TY  - CPAPER
TI  - Rhythmic Music Codec: An MDCT-Based Music Codec with Onset-Localized Block Switching and Tempo-Synchronized Prediction
AU  - Bao, Haotian
AU  - Bukovec, Nick
AU  - Krinsky, Summer
AU  - Bosi, Marina
T2  - AES Show 2026 Nashville
SP  - 1
EP  - 12
M1  - Paper 43
PY  - 2026
DA  - 2026/10/30
UR  - https://aes.org/publications/elibrary-page/?id=23474
PB  - Audio Engineering Society
CY  - Gaylord Opryland Resort
LA  - en
AB  - This paper presents the Rhythmic Music Codec (RMC), a novel audio codec that exploits the temporal structure in music. RMC builds upon a baseline codec based on the Modified Discrete Cosine Transform (MDCT) by incorporating two complementary modules: transient-aware block switching and rhythmic prediction. A multi-layer transient detector identifies onsets and triggers a sequence of variable-length analysis/synthesis windows providing enhanced time resolution in the proximity of each detected attack. A tempo-synchronized prediction module identifies rhythmically related regions in previously decoded audio and encodes the prediction residual only when doing so reduces the coding cost. Subjective listening tests following the ITU-R BS.1116 recommendation compare the baseline codec, the block switching and prediction modules individually, and the complete RMC system. The results demonstrate that the combined codec provides a statistically significant improvement in perceived audio quality over the baseline while incurring only a small, bounded increase in decoder complexity, with the added cost concentrated in the encoder.
ER  -