SCOM 7330 Speech Library Tool

This is a tool to manipulate Audio Libraries (such as CustomAudioLib.bin) for the SCOM 7330 repeater, reverse engineered from the provided BuildSpeechLib.exe.

The official tool, as well as default audio libraries, manuals, firmware, etc. can be found at http://www.scomcontrollers.com/7330firmware .

Speech Library Format

Remember that this is reverse engineered (so not guaranteed to be accurate), and subject to change with firmware updates

This was derived from firmware release V1.8b, BuildSpeechLib.exe V1.2.0 (built 5/13/2018)

The speech library file is composed of 4 sections:

• Header (0x100 bytes)
• Image Header (0x100 bytes)
• Word Index (variable length, rounded up to the nearest 0x100)
• Word Data (variable length)

The header and image header contain metadata and the word index is a lookup table into the data. Unused space is filled with 0xff instead of 0x00, for unclear reasons.

In the following sections, byte ranges are presented as inclusive:exclusive, like in Python. For example, 0x00:0x05 includes the bytes 0x00, 0x01, 0x02, 0x03, and 0x04. All addresses in tables are relative to the start of the section, unless stated otherwise.

Header

• Address: 0x00
• Length: 0x100

• name, version, and timestamp are just ASCII strings, and have no apparent function beyond display.
• mode is 3 in BuildSpeechLib.exe's normal mode, 2 in the extended arguments form. I am not sure what this means, exactly.
• firstFree is the size of the file minus 0x100, presumably to account for this header's size. Unclear what this is used for, as it also occurs (with the 0x100 included) in the Image Header
• zeros is two two-byte arguments to the function in BuildSpeechLib.exe which were passed in as literal zeros. Unclear function.

Image Header

• Address: 0x100
• Length: 0x100

• I'm not sure what the purpose of the constant in 0x00:0x03 is
• index_size_mid is the middle byte of the Word Index size. Since that size is rounded to 0x100, this is just the index size shifted down 1 byte.
• max_word + 1 is the highest word file contained, plus one. Not really sure why it's plus one.
• firstFree is the length of the file. Seems to be redundant with the value in the Header.

Word Index

• Address: 0x200
• Length: 4 times the maximum word code, rounded up to the nearest 0x100

This section is a lookup table mapping the word code to the address of the entry in the Word Data section, relative to the file start. All un-used entries are blank (0xff). Each entry is 4 bytes long, but the address is stored just in the first 3 bytes (with the last one being 0xff.

Word Data

• Address: 0x200 plus the the length of the Word Index, as stored in the Image Header.
• Length: 3 times the number of files, plus the length of each file

For each entry:

• end of file data address is the address of the end of the audio file data.
• file data is the contents of a raw file (as per Audio Files), but with every byte greater than 127 xor'ed with 127 for some unclear reason.

Audio Files

The input/output audio files are in the format specified in the SCOM manual:

• 8000 Hertz sampling rate
• Single channel (mono) audio
• μ-law encoding
• Raw headerless file

You can convert wav files (for example) to this format with sox:

sox --type wav <input file>.wav --type ul --rate 8k <output file>.raw

or from this format to wav:

sox --type ul --rate 8k --channels 1 <input file>.raw --type wav <output file>.wav

Disclaimer

I am not affiliated with S-COM in any way. I make no guarantees of the correctness of the code. Don't break your stuff.