Audio Codec Mode Selection Using Estimated Quality Measures

The patent applies preliminary action by estimating quality measures (segmental SNR) for different encoding algorithms before actual encoding occurs. The estimator computes quality metrics in advance using simplified models, allowing the mode selector to make informed decisions without performing complete encoding/decoding cycles for each candidate algorithm, thus reducing computational complexity while maintaining quality assessment accuracy.

The patent introduces an intermediary quality estimator that mediates between the encoding algorithms and the mode selection process. This estimator acts as a surrogate that predicts encoding quality without requiring full encoding operations, using segmental SNR calculations on weighted signal domains to provide reliable quality metrics at lower computational cost.

The patent replaces the mechanical system of actual encoding/decoding operations with a computational model that estimates quality measures. Instead of physically executing the encoding process to assess quality, the system uses mathematical estimators that compute segmental SNR in weighted signal domains, substituting complex mechanical encoding operations with lighter computational calculations.

The patent changes parameters by working in the weighted signal domain rather than the original audio domain. By applying weighting functions to the signal before estimation, the system transforms the problem space to make quality assessment more efficient and accurate, allowing reliable quality prediction with reduced computational effort compared to unweighted full-encoding approaches.