Audio Signal Processing MPEG1 Audio

1
Audio Signal Processing-- MPEG-1 Audio

• Shyh-Kang Jeng
• Department of Electrical Engineering/
• Graduate Institute of Communication Engineering

2
History

• Moving Picture Expert Group (MPEG)
• Established in 1988
• Joint Technical Committee (JTC1) ISO, IEC
• Develop standards for coded representation of
  moving pictures and associated audio
• Original work items
• MPEG-1, up to 1.5 Mb/s (ISO/IEC 11172)
• MPEG-2, up to 10 Mb/s (ISO/IEC 13818)
• MPEG-3, up to 40 Mb/s
• MPEG-3 was dropped in July 92

3
History (cont.)

• MPEG-4
• First proposed in 1991
• Approved in July 1993
• Targets audiovisual coding at very low bit rates
• Scalability, 3-D, etc.
• ISO/IEC FDIS in 1999 (ISO/IEC 14496)
• MPEG-7
• Started in the Fall of 1996
• Standardize the description of multimedia
  contents of multimedia data base search
• Scheduled to become ISO/IEC standard in 2001

4
MPEG-1 Audio Goals

• Define coded representation of high quality audio
  for storage media and a method for decoding high
  quality audio signals
• Input of the encoder and output of the decoder
  are compatible with existing PCM standards such
  as CD and DAT
• Support one or two main audio channels
• Sampling frequencies
• 32 KHz, 44.1 KHz, 48 KHz

5
MPEG-2 Audio Goals

• Define the multichannel extension to MPEG-1 audio
  (MPEG-2 BC, backward compatible)
• Define an audio coding standard at lower sampling
  frequencies (16 KHz, 22.5 KHz, 24 KHz) than
  MPEG-1
• Define a higher quality multichannel standard
  than MPEG-1 extensions (MPEG-2 non-backward
  compatible, NBC, later MPEG-2 Advanced Audio
  Coding, AAC)

6
MPEG-4 Audio Goals

• Provide high coding efficiency
• Data rates from 2 kb/s/ch to 64 kb/s/ch
• Provide content based interactivity
• Flexible access and manipulation (e.g.,
  pitch/speed modification)
• Support synthetic audio and speech
• Structured audio (SA)
• Text to speech (TTS)
• Provide additional effects
• Post-processing, e.g., reverb, 3D

7
MPEG-1 Audio

• Three-part compression standard
• Compression of synchronized video and audio at a
  total data rate of 1.5 Mb/s
• Specification
• Syntax of the coded bit stream
• Decoding process
• Compliance tests for assessing the accuracy of
  the decoder
• Encoder is not specified

8
MPEG-1 Audio (cont.)

• Audio channel configurations
• Monophonic mode for a single channel
• Dual monophonic mode for two independent channels
• Stereo mode in which the two channels are sharing
  bits
• Joint-stereo mode
• Data rates
• 32 kb/s/ch to 224 kb/s/ch (depending on the
  sampling rate, compression ratios up to 241)

9
MPEG-1 Audio Layers

• Layer I
• Simplest configuration, 32 to 224 kb/s/ch
• Best for data rates above 128 kb/s/ch
• Used in Philipss DCC at 192 kb/s/ch
• Layer II
• Intermediate complexity, 32 to 384 kb/s/ch
• Best for data rates of 128 kb/s/ch
• Used in DAB, CD-Interactive, etc.
• Layer III
• Highest quality and complexity, 32 to 160 kb/s/ch
• Best for data rates below 128 kb/s/ch
• Used for transmission over ISDN, Internet, etc.

10
MPEG-1 Audio Layers (cont.)

• Single-chip, real-time decoders exist for all
  three layers
• Layers II and III
• Perceptually lossless at 128 kb/s/ch (compression
  ratio of 61, 16 bits per sample, 48 KHz sampling
  rate)
• Selected by ITU-R TG 10/2 for broadcast
  applications

11
MPEG-1 Encoder Building Blocks
32 sub-bands (Layers I, II) 576 sub-bands (Layer
III)
12
MPEG-1 Decoder Building Blocks
13
MPEG-1 Filter Bank

• A 512-Tap OTDAC filter bank is common to all
  layers
• 32 filters subdivide audio signals onto 32 equal
  width bands
• At low frequencies a single sub-band covers
  several critical bands
• The critical band with the highest SMR dictates
  the number of quantization bits needed for the
  entire sub-band
• Small error (in absence of quantization, the
  total ripple is

14
MPEG-1 Filter Bank (cont.)

• Layer III filter bank
• Cascaded with a 36-point MDCT for a total of 576
  frequency lines in steady state conditions
• A 12-point MDCT is used for transients
• Frequency resolution at 48 KHz
• Layers I and II 750 Hz
• Layers III 42 Hz
• Time resolution at 48 KHz
• Layers I and II 0.66 ms
• Layer III 4 ms

15
MPEG-1 Psychoacoustic Models

• Two models
• Model 1 (less complex)
• Model 2 (includes specifics to support Layer II)
• Either works for all layers
• For low level of compression the psychoacoustic
  models can be bypassed
• Perform a separate analysis of the input signal
  via a Hanning windowed FFT
• Model 1 512 points for Layer I, 1024 for Layers
  II and III
• Model 2 1024 points for all layers

16
MPEG-1 Psychoacoustic Models (cont.)

• The analysis needs to be time-aligned with the
  main time to frequency mapping of the input
  signal
• Model 1 provides one evaluation per frame
• Model 2 provides two evaluations per frame
• The first centered on the first half of the main
  path data
• The second centered on the second half of the
  main path data
• The highest SMR is chosen

17
MPEG-1 Psychoacoustic Models (cont.)

• Separation of noise like vs. tone like signal
  components
• Model 1 based on local peaks the remaining
  spectral values are added into a non-tonal
  component per critical band
• Model 2 use data from the two previous windows
  to predict the component values for the current
  window
• Spreading of masking through adjacent bands

18
MPEG-1 Layers Frames
19
Bitstreams for MPEG-1 Audio Layers
20
MPEG-1 Layer I

• Frame size
• 12 samples/subband 32 subband 384 samples
  (corresponds to 8 ms at 48 KHz)
• 1 scale factor and 1 bit allocation for each 12
  samples sub-block
• Bit allocation
• 4 bits per allocation (values 0, , 15)
• 432 128 bits per channel
• Scale factors
• 6 bits per scale factor
• Up to 632 192 per channel

21
MPEG-1 Layer II

• Frame size
• 312samples/subband32 subband 1152 samples
  (corresponds to 24 ms at 48 KHz)
• 1 bit allocation and up to 3 scale factors for
  each 36 subband samples
• Bit allocation
• 2 to 4 bits per subband (more bits for lower
  subband) totaling 26 to 188 bits/frame
• Scale factors
• Up to 6323 576 bits per channel

22
MPEG-1 Layer III

• Frame size
• Same as Layer II, 1152 samples
• Non-uniform quantization
• Raises input to 3/4 power (the decoder
  relinearizes the values by raising them to 4/3),
  i.e., bigger values are with less accuracy
• Scale factors bands cover approximately critical
  bandwidths
• Dynamic noise allocation

23
MPEG-1 Layer III (cont.)

• Entropy coding
• At high frequencies a series of zeros is coded by
  run length coding
• Count 1 region codes values less than 1 in
  absolute values with 4-D Huffman coding
• Big values region covers the remaining values
  by 2-D Huffman coding
• Bit reservoir
• Uses a 9-bit pointer pointing to the starting
  byte for that frame
• Buffer length 7680

24
Stereo Coding

• Two types
• Layers I, II, and III Intensity stereo coding
• Layer III Mid/Side (M/S) Coding
• Intensity stereo
• Upper frequencies subbands are coded with a
  single summed signal and a scale factor for each
  channel subband is transmitted
• M/S stereo
• In certain frequency ranges the sum or difference
  are coded.