AMR

Adaptive Multi-Rate (AMR) is the new speech codec for GSM and UMTS. There are two different versions of AMR. The basic version, also called AMR-Narrowband (AMR-NB), which is mainly intended for use by GSM and AMR-Wideband (AMR-WB), which is mainly intended for use by UMTS. Wideband AMR uses a speech bandwidth of 50 - 7000 Hz, whereas the bandwidth of narrowband AMR is 300 - 3400 Hz. This gives wideband AMR a more natural speech quality.

AMR-Narrowband
The narrowband AMR speech codec is actually a combination of eight speech codecs with bit rates of 12.2, 10.2, 7.95, 7.4, 6.7, 5.9, 5.15 and 4.75 kbps. The Narrowband AMR speech codec utilises the algebraic code excitation linear prediction (ACELP) algorithm that is employed also in GSM EFR and D-AMPS EFR codecs. The six lowest bit rates can be used in the half-rate channel mode.

The speech codec modes perform also error correction and bad-frame detection. This is done by a combination of convolutional coding for the error correction and a cyclic redundancy check (CRC) for the detection of bad frames.

AMR-Wideband
AMR-Wideband has nine different codec rates, ranging from 6.60 to 23.85 kbps. The codec rates from 6.60 to 19.85 kbps can be supported by GSM as well. AMR-Wideband uses the same algorithm as AMR-Narrowband.

The AMR wideband codec is also known as HD Voice.

Choosing the right mode
The network dynamically chooses the AMR speech codec mode and the full rate or half rate channel mode for each call. The choice is based on measurement of the received quality by the mobile station and the base station and on the network load. At high traffic loads the network uses AMR Half Rate extensively. When the network is less busy, it assigns AMR Full Rate coding to as many calls as possible, starting with those experiencing the poorest radio conditions. The network also chooses the best error correction level within AMR Full Rate and AMR Half Rate to achieve the best call quality. This process, known as codec mode adaptation, results in improved voice quality throughout the cell and increases overall coverage.

The codec mode chosen in the uplink may be different from the one used in the downlink direction, but the channel mode (HR or FR) must be the same.

AMR also contains voice activity detection and discontinuous transmission (VAD/DTX). These are used to switch off the encoding and transmission during periods of silence, thereby reducing radio interference and extending the battery lifetime.