Speech and Audio Coding

1
Speech and Audio Coding

• Reference
• Voice Compression and Communications
• Principles and Applications for Fixed and
  Wireless Channels by Lajos Hanzo, F. Clare A.
  Somerville, Jason P. Woodard

PCM,DPCM, ADPCM
CELP,VSELP
LPC
2

• MOS (Mean Opinion Score) subjective measure
• 1 bad, 2-poor, 3-fair, 4-good, 5-execellent
• PCM G.711, ADPCM G.726/G.723
• CELP G.728
• Sub-band coding

• Speech Coding Objectives
• High perceived quality
• High measured intelligibility
• Low bit rate (bits per second of speech)
• Low computational requirement (MIPS)
• Robustness to successive encode/decode cycles
• Robustness to transmission errors
• Objectives for real-time only
• Low coding/decoding delay (ms)
• Work with non-speech signals (e.g. touch tone)

3
Chap 1. Quantization (PCM)

• Introduction
• Discrete in Time
• Nyquist sampling theorem,
• Anti-aliasing
• Discrete in Amplitude
• ADC (uniform quantizer)
• signal amplitude ? -V, V
• - Quantization characteristic
• - Quantization level L, step q
• - signal power

Midrise
Midtread
4
- Quantization Error (Noise) granular
distortion only (vs. overload distortion)
- SNR (Signal to Quantization noise
ratio) where R is the
quantization bit number
5

• Rate distortion Theorem
• - lower distortion need higher data rate
• From Information Theory
• If x (signal) is Gaussian distributed
• SNR of other signal distributions
• - same S.D. , uniform distribution had smallest
  dynamic range
• ? uniform distributed signal has smallest SNR

6

• Optimal Quantization level for non-uniform
  distributed signal-
• highly distributed signal range
• ? small quantization error
• ? non-uniform quantization step

• Of course, we can find the optimal quantization
  level and SNR when signal pdf is given.

• Companding
• - using a compressor to transform the signal pdf
  into uniform distribution.
• - compressor y c(x)
• - from Probability Theory

7
(No Transcript)
8

9

10

• Approximation of c(x)

• ?-law (United State standard)

11

• A-law (Europe standard)

12

• Piecewise linear approximation of A-law - 16
  segments

13

• Design of optimal non-uniform quantizer
• (Lloyd-Max quantizer)

14

• Step 1 of iterative method

• Step 2 of iterative method

15

• Iterative method (repeat Step 1, 2)
• ? converge to optimal solution
• Initial value ? uniform quantizer

16

• Simulation results (form ref10)