Quantization with Signal Adding Noise Shaping using Long-range Look-ahead Optimization

摘要

A digital audio signal re-quantization approach using noise shaping through the addition of extra signals is proposed, and the features and performance characteristics of this approach, which were previously proposed by the author but not closely studied, are investigated. It is found that the noise shaping performance is slightly better than that of the conventional approach and the perceptual evaluation is superior in terms of the sound source image fineness, particularly when a wide optimization horizon is used in the additional signal calculation. As a wide horizon requires a high computational load, an optimization pruning scheme is proposed to reduce the calculation time, which is evaluated experimentally.