CELP

CELP stands for code-excited linear prediction. It refers to any of a class of algorithms used in speech codecs to "tune out" background noise by focusing on the expected frequencies of speech.

Codecs that use any type of CELP:

• MPEG-4 (optional for audio)
• Opus
• Speex
• G.728

Further reading

• https://people.xiph.org/~jm/papers/aes120_speex_vorbis.pdf
• https://www.speex.org/docs/manual/speex-manual/node9.html
• CELT (at Wikipedia) - CELT was merged into Opus