Formats

There appear to be [at least] two file formats for AMR codec data. These are:

1. ".AMR" Files stored in "AMR File Storage Format". This is specified in draft-ietf-avt-rtp-amr-10.txt, Sec. 6.2, which is included with the Ericsson AMR converter tool. (The draft is an early form of RFC 3267). These are handled by the Ericsson AMR tool, as well as by Nokia Series 60 phones. They have a header of "#!AMR\n".

2. ".COD" Files stored in "AMR Interface Format 2." This is specified in 3GPP TS 26.101, Appendix A. These are coded and decoded by the 3GPP TS 26.104 floating point reference codec source package.

Tools

• amrConv.py

To aid development work with these two formats (and to clarify their differences) I have written amrconv.py, a small Python program that converts between the two formats. Here is the Usage message:

This program converts between:
  .amr files [AMR File Storage Format (draft-ietf-avt-rtp-amr-10.txt, Sec. 6.2)]
        and
  .cod files [AMR Interface Format 2 (3GPP TS 26.101, Annex A)]

Usage: amrconv.py filename.ext
        When .ext == '.amr' creates filename.ext.cod file
        When .ext == '.cod' creates filename.ext.amr file

Differences

• Have a 6 byte header: "#!AMR\n".
• Take two nibbles (8 bits) to express the frame flags [Format: "PTTTTVPP", where P is "Pad" (0), T is "Frame Type", and V is "Valid"].
• Pack the bits of the frame into bytes in big-endian order, that is, the most significant bit (0x80) of each byte is actually the first bit of the byte, the least significant bit (0x1) is the last bit.

• Have no header.
• Require a single nibble to express the frame flags: "TTTT", which is the Frame Type field. The actual speech frame bits begin directly after this, in the 2nd nibble of the first byte of the frame.
• Pack the bits of the frame into bytes in little-endian order. This is the way programmers usually think of packing bits in bytes. The least significant bit is the first bit of the byte, followed by the 0x2 bit, the 0x4 bit, etc.

Other Convertors A person named Xabi has a created a AMR <-> IFS converter (in perl).