摘要 |
For facilitating the transmission or recording of digital audio signals in a narrow band channel or at a reduced data rate without deterioration of audio quality detectable by the human ear, a digital audio signal is subdivided into overlapping time window segments and the segments are subjected to fast Fourier transformation, with generation of scaling factors. The spectral values resulting from the transformation are converted into separate magnitude and phase values, the former being logarithmically quantized and the latter being linearly quantized. The magnitude values corresponding to an upper portion of the spectrum are collected into groups of a size that increases towards the high frequency end of the spectrum and the magnitude values of each group are rms averaged, so that one value can serve in place of several magnitude values, thereby reducing the data to be transmitted or recorded. A final multiplexing of processed magnitude and phase values and of scaling factors precedes transmission or recording. For recovery of the digital audio signal there is corresponding preliminary demultiplexing operation. The magnitude values from the demultiplexer and the corresponding phase values are then used to provide spectral values each expressed in a real part and an imaginary part. Those outputs and the scaling factors are then subjected to inverse fast Fourier transformation (FFT), followed by inverse windowing, to produce an approximated audio digital signal which, when reproduced, cannot be distinguished by the human ear from a reproduction of the original digital audio signal.
|