MPEG-1, Layer I and II

The ISO MPEG-1 audio coding standard, described in [8], consists of three different layers (I, II and III). The layers differ quite a bit in coding and psychoacoustic models.

Layer I and II both use only a QMF filterbank of order 511, which has a 96 dB rejection of sidelobes and a steep pass-to-stop-band transition. The filterbank splits the input data in 32 subbands.

The psychoacoustic model consists of a 512 point FFT (1024 for layer II), from which a masking threshold is calculated. Peaks of the spectrum are considered tonal components and treated differently from the remaining noise components. The tonal and noise masking thresholds are added, to produce a masking threshold. This is used together with a linear quantizer to achieve appropriate rate-distortion performance in each subband, so that each subband has the same noise-to-mask-ratio and the bits add up to the preferred bit-rate. The quantized subband data is fixed-length encoded.

MPEG-1 Layer I at 384 kbit/s is used in Philips Digital Compact Cassette (DCC), and Layer II at 256 kbit/s is used for Digital Broadcast Audio (DBA).