Verfahren und Vorrichtung zum Verarbeiten von zeitdiskreten Audio-Abtastwerten

摘要

In order to obtain an integer transformation, which provides integer initial values, the TDAC (time domain alaising cancellation) function of an MCDT (modified discrete cosine transform) is executed explicitly in the time domain before the forward transformation. In the event of overlapping windows, this leads to a Givens rotation that can be represented by lifting matrices. Time-discrete sampled values of an audio signal are initially combined in pairs to form a vector in order to be sequentially subjected to the action of a lifting matrix after the other. According to the invention, a rounding step (20, 24, 28) is executed by using a lifting matrix after each multiplication (18, 22, 26) of a vector whereby resulting only in integers on the output side. A spectral representation with integer spectral values can be obtained by transforming (14) the windowed integer sampled values with an integer transformation. The inverse imaging with an inverse rotation matrix and with corresponding inverse lifting matrices leads to a precise reconstruction. The inventive concept provides a loss-free transformation, which can be directly coupled to an entropy coder without quantization in order to obtain a windowing method and transformation method that can be implemented advantageously with regard to hardware.