ETSI - TS 126 104
Digital cellular telecommunications system (Phase 2+) (GSM); Universal Mobile Telecommunications System (UMTS); LTE; ANSI-C code for the floating-point Adaptive Multi-Rate (AMR) speech codec
Organization: | ETSI |
Publication Date: | 1 April 2017 |
Status: | inactive |
Page Count: | 26 |
scope:
This Technical Standard (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 3GPP TS 26.110, or in packet-based (e.g., H.323) applications. The bit-exact fixed-point ANSI-C code in 3GPP 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 specification is the only standard conforming non-bit-exact implementation of the Adaptive Multi Rate speech transcoder (3GPP TS 26.090 [2]), Voice Activity Detection (3GPP TS 26.094 [6]), comfort noise generation (3GPP TS 26.092 [4]), and source controlled rate operation (3GPP TS 26.093 [5]). The floating-point code also contains example solutions for substituting and muting of lost frames (3GPP TS 26.091 [3]).
The fixed-point specification in 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 specification 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 specification 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.