Computing, IT & AI

Linear predictive coding and CELP speech coding

1967-1985 CEBishnu S. Atal (Bell Labs), with Manfred SchroederAtal & Hanauer (1971); Schroeder & Atal, CELP (ICASSP 1985)Diaspora (work abroad)

1967

Methods for compressing speech by predicting each sample from the ones before it; the basis of mobile phone voice codecs.

Dated: 1967–1985
As scholars put it: 1967-1985 CE
Originator: Bishnu S. Atal (Bell Labs), with Manfred Schroeder
Source text: Atal & Hanauer (1971); Schroeder & Atal, CELP (ICASSP 1985)
Region: Diaspora (work abroad)
Discipline: Computing, IT & AI
Evidence: Documented
Confidence: High
Verification: Verified-web

Background

Bishnu Atal, born in Kanpur and educated at the University of Lucknow and IISc, joined Bell Labs in 1961. From the late 1960s he developed linear predictive coding of speech, which models the vocal tract as a filter. In 1985 he and Manfred Schroeder introduced code-excited linear prediction (CELP), which gave good speech quality at very low bit rates.

Significance

CELP and its variants are the most widely used speech coding algorithms. Almost every mobile phone call compresses speech with a descendant of this work.

Fair to say

Linear prediction was also developed independently in Japan (Itakura and Saito).

Path to modern innovation

Voice codecs in 2G-5G mobile networks and in VoIP.

Global impact

Mobile phone voice codecs.
Speech coding in MPEG-4 audio.

References

3Code-excited linear prediction (CELP): high-quality speech at very low bit rates M. R. Schroeder and B. S. Atal, 1985