Adaptive Multi-Rate

Description

The Adaptive Multi-Rate (AMR) codec is a family of speech coding algorithms designed for mobile communication systems. It operates by compressing digitized speech signals into a range of bit rates, from 4.75 kbps to 12.2 kbps for narrowband AMR, and 6.60 kbps to 23.85 kbps for AMR-Wideband (AMR-WB). The core principle is source-controlled rate adaptation, where the codec mode (bit rate) is selected based on real-time assessments of radio channel quality and network load. This selection is managed by the network through in-band signaling or explicit control messages, allowing the system to prioritize either voice quality or capacity as needed.

Architecturally, AMR integrates into the voice processing chain at the User Equipment (UE) and the core network's Media Gateway (MGW) or Media Resource Function (MRF). The codec employs Algebraic Code-Excited Linear Prediction (ACELP) for narrowband and a modified ACELP for wideband, providing efficient modeling of the vocal tract and excitation signal. Key components include the speech encoder/decoder (codec), a Voice Activity Detector (VAD) for discontinuous transmission (DTX), and a comfort noise generator (CNG) to mask transmission gaps. The codec operates in conjunction with channel coding and interleaving in the physical layer to ensure robustness against errors.

In the network, AMR plays a pivotal role in the voice bearer path. For circuit-switched voice in GSM and UMTS, it is applied directly over the air interface, with the TRAU (Transcoder and Rate Adaptation Unit) or MGW handling rate adaptation and transcoding to/from PSTN codecs like G.711. For Voice over LTE (VoLTE) and Voice over NR (VoNR), AMR-WB is used as the primary codec within the IP Multimedia Subsystem (IMS), encapsulated in RTP/UDP/IP packets. The codec's adaptability allows the Radio Access Network (RAN) to command a lower bit rate during poor radio conditions, increasing channel coding protection and maintaining call continuity, or a higher bit rate for superior quality in good conditions.

The evolution to AMR-WB (marketed as HD Voice) extended the audio bandwidth from 300–3400 Hz to 50–7000 Hz, significantly improving naturalness and intelligibility. This required updates throughout the voice chain, including acoustic components in devices and wider bandwidth support in networks. AMR's design also facilitates seamless handovers between different radio access technologies (e.g., GSM to UMTS) through codec mode renegotiation and transcoding in the core network. Its standardized frame structure and control mechanisms ensure interoperability across vendors and operators, making it a foundational technology for global mobile voice services.

Purpose & Motivation

AMR was created to address the limitations of fixed-rate speech codecs used in early digital mobile systems like GSM, which employed the Full-Rate (FR) and Half-Rate (HR) codecs. These fixed codecs could not adapt to varying radio conditions: the FR codec provided consistent but sometimes inadequate quality under interference, while the HR codec offered higher capacity but lower quality. The primary motivation was to optimize the trade-off between voice quality and spectral efficiency dynamically, allowing networks to maintain acceptable quality during congestion or poor coverage while maximizing capacity during ideal conditions.

Historically, the introduction of AMR in 3GPP Release 99 (for GSM) and its adoption for UMTS was driven by the need for a unified, robust voice codec that could leverage advancements in digital signal processing. It solved the problem of inefficient spectrum usage by enabling the network to command a lower bit rate (and thus stronger error protection) during cell edge or interference scenarios, reducing drop calls. Conversely, in clear conditions, it could use higher bit rates for near-toll-quality audio. This adaptability was crucial for supporting higher user densities and improving overall service reliability.

Furthermore, AMR provided a migration path for enhanced voice services. It laid the groundwork for wideband audio (AMR-WB), which addressed the growing demand for high-definition voice experiences as networks evolved to packet-switched IMS architectures in LTE and 5G. By standardizing a single, adaptable codec family, 3GPP ensured backward compatibility and smooth interworking between legacy and modern networks, reducing complexity for operators and enabling consistent voice quality across generations.

Key Features

• Dynamic bit rate adaptation from 4.75 to 12.2 kbps (narrowband) based on channel conditions
• Wideband audio support (AMR-WB) with 50–7000 Hz bandwidth for HD Voice
• Source-controlled rate selection managed by network for optimal quality/capacity trade-off
• Integrated Voice Activity Detection (VAD) and Discontinuous Transmission (DTX) for power saving
• Robustness against transmission errors through link adaptation and channel coding synergy
• Standardized frame structure enabling multi-vendor interoperability and transcoding

Evolution Across Releases

Defining Specifications