Block diagrams of Layer I and Layer II encoders are given in Figure .1. An analysis filterbank splits the input signal with sampling rate F, by dividing it into 32 equally spaced subband signals with sampling rate F32. In each of the 32 subbands, 12 consecutive samples are assembled into blocks with the equivalent of 384 input samples. All of the samples within one block are normalized by a scale factor so that they all have absolute values less than one. The choice of a scale factor is done by first finding the sample with the maximum absolute value, and then comparing it to a scale factor table of 63 allowable values. After normalization, samples are quantized and coded under the control of a psychoacoustic model. Detailed psychoacoustic analysis is performed through the use of a 512 (Layer I) or 1024 (Layer II) point fast Fourier transform (FFT) in parallel with the subband decomposition. The bitallocation unit determines the quantizer resolution according to the targeted bit rate and the perceptual information derived from the psychoacoustic model. Layer II introduces further compression with respect to Layer I through three modifications. First, the overall information is reduced by removing redundancy and irrelevance between the scale factors of three adjacent 12-sample blocks. Second, a quantization table with improved precision is provided. Third, the psychoacoustic analysis benefits from better frequency resolution because of the increased FFT size.
đang được dịch, vui lòng đợi..