Method and apparatus for compressing a digital input signal with block floating applied to blocks corresponding to fractions of a critical band or to multiple critical bands

摘要

A method for compressing a digital input signal to provide a compressed signal in which spectral components are derived from the digital input signal and are divided into critical bands. The spectral components at lower frequencies are also divided into a block having a bandwidth corresponding to that of plural critical bands. Block floating processing is applied to the block to generate a floating coefficient. Finally, the spectral components in each critical band are quantized using an adaptively-allocated number of bits. Additionally, the word length indicating the number of bits allocated for quantizing the spectral components in each critical band is included in the compressed signal together with the floating coefficient for the block as the only floating coefficient for the critical bands in the block. This reduces the number of bits in the compressed signal. The spectral components in a higher-frequency critical band are divided into blocks having a bandwidth that is narrower than the bandwidth of the critical band. Block floating processing is applied to the blocks to generate a floating coefficient for each block. Finally, the spectral components in each block are quantized using an adaptively-allocated number of bits. The word length for each block is included in the compressed signal together with, instead of the floating coefficient for each block, the allowed noise level for the critical band.