What is Adaptive Multi-Rate (AMR)?

Adaptive multi-rate compression ( AMR ) is a patent that optimizes speech coding. AMR was selected by the standard speech coding 3GPP in October 1998 and is currently widely used in GSM and UMTS. It uses 1-8 different bit rate encodings.

AMR is also a file format that uses AMR codecs to store voice and audio. Many modern mobile phones handsets can store short recordings in AMR format, and there are free and proprietary programs (see software support) to convert between this and other formats, although AMR is a speech format and is unlikely Provides ideal result audio for other formats. Common file extensions are .amr. AMR also has another storage format, which is suitable for applications with more advanced storage format requirements, such as random access or synchronization with video. This format is a 3GP container format specified by 3GPP based on the ISO base media file format. ^[1]

In order to guarantee a sampling rate of 160 sampling points every 20 milliseconds, AMR uses different technologies such as ACELP, DTX, VAD, and CNG, and chooses the best coding mode to meet the requirements of local channel and channel frequency bands. If the broadcast conditions are poor, the source coding decreases and the channel coding increases. This can improve the quality of the network connection, but at the expense of speech intelligibility. In special cases the AMR increases by approximately S / N = 4-6 dB.

Figure 1 lists a total of 14 types of AMR codes, including 8 full-frequency (FR) and 6 half-frequency (HR).

• The sampling rate is 8 kHz / 13-bit (160 sampling points every 20ms), and only the signals in the range of 200-3400 Hz are retained after filtering.

• The encoder uses 8 bit rates: 12.2, 10.2, 7.95, 7.40, 6.70, 5.90, 5.15, and 4.75 kbit / s.

• The generated frame length is 95, 103, 118, 134, 148, 159, 204, or 244bit, and the corresponding bit rate is 4.75, 5.15, 5.90, 6.70, 7.40, 7.95, 10.2, or 12.2 kbit / s.

• AMR uses Discontinuous Transmission (DTX), Voice Activity Detection (VAD), and Comfort Noise Generation (CNG) to reduce bandwidth during silent periods.

• The algorithm delay is 20ms per frame. When the bit rate is 12.2, there is no look-ahead delay. The look-ahead delay for other rates is 5 ms. Note that a pseudo-look-ahead delay of 5 ms allows accurate frame-width mode conversion of the remaining rates.

• AMR is a mixed language encoder using Algebraic Code Excited Linear Prediction (ACELP).

• If the complexity of the algorithm is 5, then

  The AMR codec contains multiple patents from Nokia, Ericssion, NTT and VoiceAge, the last one being the License Administrator of the AMR patent pool. VoiceAge also accepts filing patents to determine its possible need for these standards. However, since the inventors (and their lawyers) did their best to hide the patents related to AMR / AMR, it is difficult to determine whether there is a so-called invention patented WB technology related to the AMR / AMR-WB codec. Obviously, all of these patents are hidden in all other studies and general audiences who may find the prior art in the patent claims filed by patent holders of AMR / AMR-WB codecs.

  The initial fee for professional content creation tools and "Live Channel" products is $ 6,500. The minimum annual royalty is $ 10,000 and does not include initial costs in the first year. The license fee per channel has been reduced from US $ 0.99 to US $ 0.50, up to US $ 2 million per year.

  In the personal computer product category, such as media players, AMR decoders are freely licensed. The license cost for selling the encoder ranges from $ 0.40 to $ 0.30, up to $ 300,000 per year. Minimum annual royalties do not apply to licensed products that fall into the personal computer product category and use only free decoders.

  AMR codec several patents:

  • VoiceAge licensing information, including codec prices.

  • 3GPP legal issues

  • The 3G Patent Platform and its licensing policy

  • AMR Codecs as Shared Libraries-amrnb and amrwb development sites, shared libraries.