What is AMR?
AMR (Adaptive Multi-Rate) was developed in 1999 by the European Telecommunications Standards Institute (ETSI) for GSM mobile phones. It's specifically optimized for encoding human speech and has become the standard speech codec for cellular systems including GSM and UMTS networks.
Key Features
AMR is designed with voice communications in mind:
- Optimized for human speech, not general audio content
- Very efficient compression at extremely low bitrates (down to 4.75 kbit/s)
- Adaptive bitrate capability to adjust to network conditions
- Low computational complexity for mobile devices
- Narrow frequency range focused on human voice (300 Hz to 3400 Hz)
When to Use AMR
AMR is particularly well-suited for:
- Voice recordings where small file size is critical
- Voice memos on mobile devices
- Speech archives where audio quality isn't the primary concern
- Voice messaging applications
- Mobile telephony applications
Variants
There are two main AMR codecs:
- AMR-NB (Narrowband): The original and most common variant, sampling at 8 kHz for telephone-quality speech
- AMR-WB (Wideband): An enhanced version sampling at 16 kHz for better speech quality, also known as G.722.2
Bitrates and Modes
AMR-NB operates at multiple bitrates, using different modes depending on conditions:
- Mode 0: 4.75 kbit/s
- Mode 1: 5.15 kbit/s
- Mode 2: 5.90 kbit/s
- Mode 3: 6.70 kbit/s
- Mode 4: 7.40 kbit/s
- Mode 5: 7.95 kbit/s
- Mode 6: 10.2 kbit/s
- Mode 7: 12.2 kbit/s (highest quality)
Limitations
Important considerations when using AMR:
- Speech-only design: Not suitable for music or general audio content
- Mono only: Only supports single-channel (mono) audio
- Limited frequency range: Only captures frequencies up to 4 kHz (AMR-NB)
- Compatibility: Not as widely supported as MP3 or AAC outside mobile devices
- Quality: Noticeably lower quality than modern audio codecs
Technical Details
AMR files use the .amr extension and have the MIME type "audio/amr". AMR-NB is sampled at 8 kHz with a single audio channel. The format uses Algebraic Code Excited Linear Prediction (ACELP) technology and incorporates voice activity detection and comfort noise generation features for efficient transmission.
AMR vs Other Audio Formats
When compared to general-purpose audio codecs:
- Much smaller file size than MP3 or AAC for voice content
- Lower audio quality for music or non-speech content
- Less widely supported in consumer applications
- More efficient specifically for speech at very low bitrates
Conclusion
AMR remains important in mobile communications where bandwidth efficiency for voice is critical. While not suitable for music or high-quality audio applications, it's an excellent choice for speech recordings where minimizing file size is the primary concern. Its continued use in mobile networks worldwide ensures its relevance despite the availability of newer audio codecs.