georgid/AlignmentDuration

Issues

• How about the alignment effect of Jingju Mandarin? Is there a demo for testing?

  #62 opened 6 years ago by osmboy
  0

• Implement viterbi in c++

  #58 opened 7 years ago by georgid
  2

• handle non-unicode characters

  #61 opened 7 years ago by georgid
  8

• refactor FeatureExtractor.loadMFCCs

  #53 opened 7 years ago by georgid
  0

• Reduce code by refining LyricsWithModels

  #49 opened 7 years ago by georgid
  0

• replace logic for taking intervals of onsets with mir_eval.adjust_intervals

  #38 opened 8 years ago by georgid
  0

• with_section_annotations = 0, sectionLInks do not work

  #30 opened 9 years ago by georgid
  0

• how to handle words not in dictionary

  #55 opened 7 years ago by georgid
  0

• deprecate for-segments branch. test on makam. integrate in mir_eval

  #33 opened 7 years ago by georgid
  0

• reduce code not used

  #59 opened 7 years ago by georgid
  1

• give section link number as arg instead of TextGridTs and so fromTextGrid

  #39 opened 8 years ago by georgid
  0

• test WITH_SECTION_ANNO = False with new MLP model

  #41 opened 8 years ago by georgid
  0

• have the same special token for silence in all phoneme sets. Now it is REST for Mandarin and '' for English

  #50 opened 7 years ago by georgid
  1

• reduce essentia code for loading the audio

  #54 opened 7 years ago by georgid
  1

• reduce dependency on essentia

  #60 opened 7 years ago by georgid
  1

• reduce dependency on sklearn by copying code for normalization

  #57 opened 7 years ago by georgid
  0

• reduce dependecy on htkmfc

  #44 opened 8 years ago by georgid
  1

• reduce dependency on package for Levenshtein distance

  #56 opened 7 years ago by georgid
  0

• remove hard coded sampling rate

  #51 opened 7 years ago by georgid
  1

• show lyrics player on recordings with only singing voice

  #42 opened 7 years ago by georgid
  2

• missing words from dictionary

  #45 opened 7 years ago by georgid
  1

• reduce code redundancy for computing timstamps

  #52 opened 7 years ago by georgid
  0

• Move this file to dir for_english

  #48 opened 7 years ago by georgid
  0

• optimize code for expnasion to syllables

  #47 opened 7 years ago by georgid
  0

• Add Mandarin Char to mandarinSyllable class

  #46 opened 7 years ago by georgid
  0

• reducnant Lyric() constructor call

  #43 opened 8 years ago by georgid
  1

• reduce dependency on htk and scikit learn

  #40 opened 8 years ago by georgid
  1

• concatenate textGrid data.

  #32 opened 8 years ago by georgid
  3

• merge _constructTimeStampsForToken and _constructTimeStampsForTokenDetected

  #37 opened 8 years ago by georgid
  0

• when WITH_SHORT_PAUSES = 1

  #36 opened 8 years ago by georgid
  1

• merge SectionLinkMakam.loadSmallAudioFragment and SectionLinkMakam.loadSmallAudioFragmentOracle

  #35 opened 8 years ago by georgid
  0

• means, covars, weights

  #34 opened 8 years ago by georgid
  1

• detectedTokenList with flag DETECTION_TOKEN_LEVEL ='words' has Word object and not the string. So there is a problem at json.dump() in LyircsAligner

  #31 opened 8 years ago by georgid
  1

• improve roganization of MusicXML Parser: use ScoreSection class in MusicXMLParser

  #29 opened 9 years ago by georgid
  0

• reduce code of constructTransMAtrix

  #28 opened 9 years ago by georgid
  0

• - put reading pitch as input in lyricsalign not in align.FeatureExtractor.loadMFCCs

  #27 opened 9 years ago by georgid
  0

• Doesn't install if cython not available

  #22 opened 9 years ago by alastair
  3

• when withPaddedSilence, consider creating dummy Word with one phoneme sp

  #26 opened 9 years ago by georgid
  0

• repetiion of _constructHMMNetworkParameters in DurationHMM and HMM

  #25 opened 9 years ago by georgid
  0

• refine and commit LyricAligner

  #24 opened 9 years ago by georgid
  0

• hmm.Path.Path.

  #23 opened 9 years ago by georgid
  0

• dont trim file,

  #19 opened 10 years ago by georgid
  1

• in LyricsWithModels assign small deviation to phonemes and the given by as parameter one to vowels

  #21 opened 9 years ago by georgid
  0

• integrate resynthesis from sms-tools to AlignmentDuration

  #20 opened 10 years ago by georgid
  2

• remove hard-coded logic for discriminating btw two duration distributions

  #13 opened 10 years ago by georgid
  4

• optimize: do observation prob for a given feature vector for all phonemes in alphabet, not for all phonemes in phonemeNetwork

  #18 opened 10 years ago by georgid
  0

• REDUCE CODE: check if we can replace some of the code in SymbTrPrser.syllable2LyricsOneSection with code from LyricsParsing.expandlyrics2WordList.

  #15 opened 10 years ago by georgid
  0

• make consistent. phoneme2states

  #17 opened 10 years ago by georgid
  0

• make sure finalTs of referenceScore duration < actual duration of recording

  #16 opened 10 years ago by georgid
  0

• saving the phrasesAligned is not OK for HTK-based system

  #14 opened 10 years ago by georgid
  0