phonemeRecognizerWrapper 0.1.1

Package containing one wrapper script over the Allosaurus phoneme recognition library, designed for passing the Allosaurus output data to MATLAB scripts for further analysis.

phonemeRecognizerWrapper

Package containing one wrapper script over the Allosaurus phoneme recognition library, designed for passing the Allosaurus output data to MATLAB scripts for further analysis.

Usage: phonemeRecognizer.py

This script uses the Allosaurus phoneme recognition package to extract phonemic content from audio files of human speech. This script acts as a wrapper over the allosaurus package for improved formatting and piping of data to MATLAB script getPhonemesAllosaurus.m, which consequently parses the data into MATLAB friendly formant and passes it to other scripts for analysis (specifically, vowel formant extraction).

Command structure

py phonemeRecognizer.py LANGUAGE_CODE FILES EMIT_PROB

Required Arguments:

1. LANGUAGE_CODE
   • Three characters long language code supported by the Allosaurus library. For the list of available languages, use command:
     py -m allosaurus.bin.list_lang
   • To display the phonetic inventory (list of phonemes) for a specific language, use:
     py -m allosaurus.bin.list_phone [--lang <language name>]
   • See here for more info.
   • Example options:
     • "ipa" - uses the whole available phonetic inventory for recognition (less accurate)
     • "deu" - german
     • "gsw" - swiss german
     • "fra" - french
     • "eng" - english
2. FILES
   • Semicolon delimited text string containing absolute paths to all files meant for recognition. Surround the string with apostrophes ("") if any of the paths contains spaces.
   • Example:
     "C:\sounds\sound1.wav;C:\sounds\sound2.wav"

Optional Arguments:

3. EMIT_PROB
   • Allosaurus setting that determines the phoneme emission rate of the underlying model. Higher number tells the model to produce more phonemes, smaller number vice versa.
     Center is at 1.0, and optimal range that produces comprehensive outputs is 0.8 - 1.5. If omitted, default value of 1.5 is used.

Examples

• Example usage from command line: py phonemeRecognizer.py eng "C:\sound.wav" 1.0

• Example usage from MATLAB via the [status, result] = system(command) function:
  command = 'py "C:\...\phonemeRecognizer.py" eng "C:\sounds\sound.wav;C:\sounds\sound2.wav" 1.0';

  • It is also recommended to use set PYTHONIOENCODING=utf8 before the python command to ensure proper text formantting via the standard output pipe.