| Abstract | A digital input signal is compressed to provide a compressed output signal in a manner which prevents pre-echo, a common defect of data compressors. The digital input signal is divided in time into frames, and into at least one block of data in each frame. The frames include a past frame preceding a present frame. Block floating is applied to the data in the block in each frame. Quantizing bits are adaptively allocated to the data in the block in the present frame in response to the block in the past frame. Finally, the resulting quantized data is added to the compressed output signal. In a first variation, a signal energy or an allowable noise level is determined for the block in each frame. A word length is determined for the block in each frame in response to the determined signal energy or allowable noise level of the block. The quantizing bits are allocated to the data in the block in the present frame in response to the word length for the block in the past frame. In a second variation, a signal energy in the block in each frame is determined, and quantizing bits are adaptively allocated to the data in the block in the present frame in response to an interpolation between the signal energy for the block in the present frame and for the block in the past frame. |
