Waveform algorithms:
- Sample at 8000 samples/sec
- Use prediction techniques
- G.711/G.726

Source algorithms:
- Vocoders/voice coders
- Utilise speech characteristics
- Predict human speech waveforms using speech codebook
- G.728/G.729

Pulse Code Modulation (PCM), takes 8000 samples/sec, either a/u-law

Adaptive Differential PCM (ADPCM), looks at different between current and predicted sample

Code Excited Linear Prediction (CELP) predicts the next wave form and synthesises and generates the waveform with the speakers characteristics.

G.729 uses CS-ACELP, G279A is a variation that allows for less CPU intensive processing, G279B includes VAD to elimiate transmission of 'silence' within a voice transmission.