Source code for annotations.CuedSpeech.lpckeys

# -*- coding: UTF-8 -*-
:author:   Brigitte Bigi
:summary:  Keys generation of the Cued Speech from time-aligned phonemes.

.. _This file is part of SPPAS: <>

     ___   __    __    __    ___
    /     |  \  |  \  |  \  /              the automatic
    \__   |__/  |__/  |___| \__             annotation and
       \  |     |     |   |    \             analysis
    ___/  |     |     |   | ___/              of speech

    Copyright (C) 2011-2021  Brigitte Bigi
    Laboratoire Parole et Langage, Aix-en-Provence, France

    Use of this software is governed by the GNU Public License, version 3.

    SPPAS is free software: you can redistribute it and/or modify
    it under the terms of the GNU General Public License as published by
    the Free Software Foundation, either version 3 of the License, or
    (at your option) any later version.

    SPPAS is distributed in the hope that it will be useful,
    but WITHOUT ANY WARRANTY; without even the implied warranty of
    GNU General Public License for more details.

    You should have received a copy of the GNU General Public License
    along with SPPAS. If not, see <>.

    This banner notice must not be removed.



from sppas.src.config import separators

from .keyrules import KeyRules

# ----------------------------------------------------------------------------

[docs]class CuedSpeechKeys(object): """Cued Speech keys generation from a sequence of phonemes. """
[docs] def __init__(self, keys_filename=None): """Create a new LPC instance. Load keys from a text file, depending on the language and phonemes encoding. See documentation for details about this file. :param keys_filename: (str) Name of the file with the list of keys. """ self.rules = KeyRules(keys_filename)
# -----------------------------------------------------------------------
[docs] def syllabify(self, phonemes, durations=()): """Return the key boundaries of a sequence of phonemes. Perform the segmentation of the sequence of phonemes into the syllables-structure of the LPC coding scheme. A syllable is CV, or V or C. >>> phonemes = ['b', 'O~', 'Z', 'u', 'R'] >>> CuedSpeechKeys("fra-config-file").syllabify(phonemes) >>> [ (0, 1), (2, 3), (4, 4) ] :param phonemes: (list of str) List of phonemes :param durations: (list of float) List of phoneme' durations :returns: list of tuples (begin index, end index) """ # Convert a list of phonemes into a list of key classes. classes = [self.rules.get_class(p) for p in phonemes] syll = list() i = 0 while i < len(phonemes): c = classes[i] if c in ("V", "C"): if i+1 == len(phonemes): # we reach the end of the list of phonemes but a phoneme # is still in the list. syll.append((i, i)) else: c_next = classes[i+1] if c == "V": # The current phoneme is a vowel syll.append((i, i)) else: # The current phoneme is a consonant if c_next == "V": # The next phoneme is a vowel syll.append((i, i+1)) i += 1 else: if c_next == "C": # The next phoneme is a consonant too if len(durations) == len(phonemes): if durations[i] > 0.04: syll.append((i, i)) # Ignore an isolated consonant if very short duration. else: syll.append((i, i)) else: # The next phoneme is neither a vowel nor a consonant. syll.append((i, i)) i += 1 return syll
# ----------------------------------------------------------------------- # Output formatting # -----------------------------------------------------------------------
[docs] @staticmethod def phonetize_syllables(phonemes, syllables): """Return the phonetized sequence of syllables. >>> phonemes = ['b', 'O~', 'Z', 'u', 'R'] >>> lpc_keys = CuedSpeechKeys("fra-config-file") >>> syllables = lpc_keys.syllabify(phonemes) >>> lpc_keys.phonetize_syllables(phonemes, syllables) >>> "b-O~.Z-u.R" :param phonemes: (list) List of phonemes :param syllables: list of tuples (begin index, end index) :returns: (str) String representing the syllables segmentation The output string is using the X-SAMPA standard to indicate the phonemes and syllables segmentation. """ str_syll = list() for (begin, end) in syllables: str_syll.append(separators.phonemes.join(phonemes[begin:end+1])) return separators.syllables.join(str_syll)
# -----------------------------------------------------------------------
[docs] def syll_to_key(self, phonetized_syllable): """Return the key (c, v) of a phonetized syllable. >>> syllable = "p-a" >>> lpc_keys.syll_to_key(syllable) >>> ("1", "s") # consonant "1" then vowel "s" :param phonetized_syllable: (str) the phonemes of the syllable :return: (tuple) """ return self.rules.syll_key(phonetized_syllable)
# -----------------------------------------------------------------------
[docs] def keys_phonetized(self, phonetized_syllables): """Return the keys of a phonetized syllable as C-V sequences. The input string is using the X-SAMPA standard to indicate the phonemes and syllables segmentation. >>> syllable = "e.p-a.R" >>> lpc_keys = CuedSpeechKeys("fra-config-file") >>> lpc_keys.keys_phonetized(syllable) >>> "0-t.1-s.3-0" :param phonetized_syllables: (str) String representing the keys segments :return: (str) """ c = list() for syll in phonetized_syllables.split(separators.syllables): try: consonant, vowel = self.rules.syll_key(syll) c.append(consonant + separators.phonemes + vowel) except ValueError: c.append(separators.phonemes) return separators.syllables.join(c)