Tech-invite3GPPspecsGlossariesIETFRFCsGroupsSIPABNFsWorld Map
Top   in Index   Prev   Next

TS 26.104 (SA4)
ANSI-C code for the floating-point Adaptive Multi-Rate (AMR) speech codec

ToC   3GPP‑Page   ETSI‑search   Help    
V15.0.0 (PDF)2018/06  26 p.
V14.0.0  2017/03  26 p.
V13.0.0  2015/12  25 p.
V12.0.0  2014/09  25 p.
V11.0.0  2012/09  25 p.
V10.0.0  2011/04  25 p.
V9.0.0  2009/12  25 p.
V8.0.0  2008/12  25 p.
V7.0.0  2007/06  25 p.
V6.1.0  2004/04  25 p.
V5.4.0  2004/04  25 p.
V4.5.0  2003/06  25 p.
V3.5.0  2003/04  25 p.

Rapporteur:  Mr. Usai, Paolino
See also:  AMR-related TS/TR

This TS contains an electronic copy of the ANSI-C code for a floating-point implementation of the Adaptive Multi-Rate codec. This floating-point codec specification is mainly targeted to be used in multimedia applications such as the 3G-324M terminal specified in TS 26.110, or in packet-based (e.g., H.323) applications. The bit-exact fixed-point ANSI-C code in TS 26.073 remains the preferred implementation for all applications, but the floating-point codec may be used instead of the fixed-point codec when the implementation platform is better suited for a floating-point implementation. It has been verified that the fixed-point and floating-point codecs interoperate with each other without any artefacts.

The floating-point ANSI-C code in this TS is the only standard conforming non-bit-exact implementation of the Adaptive Multi Rate speech transcoder (TS 26.090), Voice Activity Detection (TS 26.094), comfort noise generation (TS 26.092), and source controlled rate operation (TS 26.093). The floating-point code also contains example solutions for substituting and muting of lost frames (TS 26.091).

The fixed-point specification in TS 26.073 shall remain the only allowed implementation for the 3G mandatory speech service and the use of the floating-point codec is strictly limited to other services.

The floating-point encoder in this TS is a non-bit-exact implementation of the fixed-point encoder producing quality indistinguishable from that of the fixed-point encoder. The decoder in this TS is functionally a bit-exact implementation of the fixed-point decoder, but the code has been optimized for speed and the standard fixedpoint libraries are not used as such.


Here   Top