New Wideband Speech Coding Standard Set by ITU

Geneva, 31 January 2002 — The ITU has approved a new Standard for high-quality digital wideband speech encoding that will bring significant improvements in terms of interoperability, easier implementation, and improved quality, for wideband voice applications and services across a wide range of communication systems and platforms.

Several important applications are envisaged for the standard. These include: Voice over IP (VoIP) and the Internet, third generation mobile communications, PSTN high-quality audio-conferencing and business applications (both in point-to-point and multi-point situations), streaming audio and speech, ISDN wideband telephony, and ISDN video telephony and video-conferencing.

The standard, known as Recommendation G.722.2, is also referred to as the Adaptive Multi-Rate Wideband (AMR-WB) codec. It has been selected by 3GPP (Third Generation Partnership Project) as the Wideband codec for GSM and 3rd generation wireless W-CDMA applications. This marks the first time that both wireless and wireline services may be able to adopt the same codec. Pierre-Andre Probst, Chairman of ITU-T Study Group 16 notes, "The AMR-WB codec is a breakthrough in speech quality. The fact that the same codec has been adopted means that interoperation between 3G and fixed IP networks will be that much easier".

Wideband speech coding, using an audio band of 50-7 000 Hz, offers major subjective improvements in speech quality compared to traditional narrowband telephone speech (200-3 400 Hz). A bandwidth of 50 to 7 000 Hz improves the intelligibility and naturalness of speech, adds a feeling of transparent communication and eases speaker recognition. The low-frequency enhancement from 50 to 200 Hz contributes to increased naturalness, presence and comfort while the high-frequency extension from 3 400 to 7 000 Hz provides improved intelligibility.

Rosario Drogo de Iacovo, Chairman of the subcommittee responsible for the work (ITU-T Study Group 16) adds, "experts from around the world have collaborated in the definition, selection and testing of this new codec. It is truly state-of-the-art". Simão Campos-Neto, Chairman of ITU-T Working Party 3/16 (Media Coding) notes, "The adoption of the same algorithm by both Standards Organizations is the result of a closely coordinated effort. We are proud to offer a single standardized solution that can be used across several industries.

The G.722.2 algorithm utilizes the Algebraic Code Excited Linear Prediction (ACELP) technology developed by VoiceAge Corp. in collaboration with Nokia, with additional novel features for improving the quality of wideband signals. G.722.2 is a multi-rate codec consisting of nine modes with bit rates of 23.85, 23.05, 19.85, 18.25, 15.85, 14.25, 12.65, 8.85 and 6.6 kbit/s. It also includes an integrated Voice Activity Detector (VAD) in conjunction with Discontinuous Transmission (DTX) and Comfort Noise Generation (CNG) for efficient low bit rate source controlled operation in background noise. The total computational complexity of the codec is estimated at less than 40 MIPS with a data RAM requirement of 6.5 kWords and total ROM requirement below 16 kWords. The standard is described in bit-exact C-code using a set of fixed-point basic operators.