ToneProt/ToneRecognition.praat

   1 #! praat\r
   2 #\r
   3 # Construct all tone patterns and look for the one closest to the given one\r
   4 #\r
   5 \r
   6 procedure FreeToneRecognition .pinyin$ .test_word$ .exclude$ .upperRegister .freqRange .durScale .startSyllable\r
   7     # Clean up input\r
   8     if .pinyin$ <> ""\r
   9         .pinyin$ = replace_regex$(.pinyin$, "^\s*(.+)\s*$", "\1", 1)\r
  10         .pinyin$ = replace_regex$(.pinyin$, "5", "0", 0)\r
  11                 # Missing neutral tones\r
  12                 call add_missing_neutral_tones '.pinyin$'\r
  13                 .pinyin$ = add_missing_neutral_tones.pinyin$\r
  14     endif\r
  15 \r
  16     .referenceFrequency = 300\r
  17     .frequencyFactor = .referenceFrequency / .upperRegister\r
  18 \r
  19     .referenceExt$ = "pitch"\r
  20     \r
  21     # Bias Z-normalized value of the distance difference between smallest and correct\r
  22     .biasDistance = 1.1\r
  23     if index_regex(config.strict$, "[^0-9]") <= 0\r
  24                 .tmp = 'config.strict$'\r
  25                 if .tmp = 0\r
  26                         .biasDistance = 1.7\r
  27                 elsif .tmp = 1\r
  28                         .biasDistance = 1.1\r
  29                 elsif .tmp = 2\r
  30                         .biasDistance = 0.6\r
  31                 elsif .tmp = 3\r
  32                         .biasDistance = 0.3\r
  33                 endif\r
  34     endif\r
  35     \r
  36     # Debugging\r
  37     .keepIntermediates = 0\r
  38     .debug = 0\r
  39 \r
  40     # Generate reference tones\r
  41     call toneScript '.pinyin$' '.upperRegister' '.freqRange' '.durScale' Pitch_'.startSyllable'\r
  42 \r
  43     # Convert input to Pitch\r
  44     if .test_word$ <> "" and .test_word$ <> "REUSEPITCH"\r
  45         Read from file... '.test_word$'\r
  46         Rename... Source\r
  47     endif\r
  48     if .test_word$ <> "REUSEPITCH"\r
  49         select Sound Source\r
  50                 call convert2Pitch 'sgc_ToneProt.minimumPitch' 'sgc_ToneProt.maximumPitch'\r
  51                 if index(.pinyin$, "3") > 0\r
  52                         select convert2Pitch.object\r
  53                         .sourcePitch = Kill octave jumps\r
  54                         select convert2Pitch.object\r
  55                         Remove\r
  56                 else\r
  57                         .sourcePitch = convert2Pitch.object\r
  58                 endif\r
  59                 select .sourcePitch\r
  60         Formula... self*'.frequencyFactor'; Normalize Pitch\r
  61         Rename... SourcePitch\r
  62     endif\r
  63     select Pitch SourcePitch\r
  64 \r
  65     .countDistance = 0\r
  66     .sumDistance = 0\r
  67     .sumSqrDistance = 0\r
  68     .correctDistance = -1\r
  69 \r
  70     .smallestDistance=999999\r
  71     sgc_ToneProt.choiceReference$ = "empty"\r
  72     select Table ToneList\r
  73     .listLength = Get number of rows\r
  74     for .i from 1 to .listLength\r
  75         select Table ToneList\r
  76         .inFile$ = Get value... '.i' Word\r
  77         # Broken third tones are still third tones\r
  78 \r
  79         if (.exclude$ = "" or rindex_regex(.inFile$, .exclude$) <= 0) and rindex_regex(.inFile$, "[\d]") > 0 \r
  80             referenceName$ = .inFile$\r
  81             select Pitch '.inFile$'\r
  82             plus Pitch SourcePitch\r
  83             .dtw = noprogress To DTW... 24 10 yes yes no restriction\r
  84             Rename... DTW'.inFile$'\r
  85             distance = Get distance (weighted)\r
  86             # Clean up\r
  87             select DTW DTW'.inFile$'\r
  88             .countDistance = .countDistance + 1\r
  89             .sumDistance = .sumDistance + distance\r
  90             .sumSqrDistance = .sumSqrDistance + distance^2\r
  91             \r
  92                         .inFile$ = replace$(.inFile$, "9", "3", 0)\r
  93             if .pinyin$ = .inFile$\r
  94                 .correctDistance = distance\r
  95             endif\r
  96 \r
  97             if .debug > 0\r
  98                 # printline 'distance' - '.inFile$'\r
  99             endif\r
 100 \r
 101             if distance < .smallestDistance\r
 102                 .smallestDistance = distance\r
 103                 sgc_ToneProt.choiceReference$ = "'.inFile$'"\r
 104             endif\r
 105 \r
 106 \r
 107             if .keepIntermediates = 0\r
 108                 Remove\r
 109             endif\r
 110         endif\r
 111     endfor\r
 112     \r
 113     if .countDistance > 1\r
 114         .meanDistance = .sumDistance / .countDistance\r
 115         .varDistance = (.sumSqrDistance - .sumDistance^2/.countDistance)/(.countDistance - 1)\r
 116         .stdDistance = sqrt(.varDistance)\r
 117         .diffDistance = .correctDistance - .smallestDistance\r
 118         .zDistance = .diffDistance/.stdDistance\r
 119 \r
 120         if .debug > 0\r
 121             printline Match: '.pinyin$' <== 'sgc_ToneProt.choiceReference$' small='.smallestDistance' Z='.zDistance'\r
 122         endif\r
 123 \r
 124         if .zDistance < .biasDistance\r
 125             sgc_ToneProt.choiceReference$ = .pinyin$\r
 126             .smallestDistance = .correctDistance\r
 127         endif\r
 128     endif\r
 129     \r
 130 \r
 131     # Clean up\r
 132     for .i from 1 to .listLength\r
 133         select Table ToneList\r
 134         .inFile$ = Get value... '.i' Word\r
 135 \r
 136         if (.exclude$ = "" or rindex_regex(.inFile$, .exclude$) <= 0) and rindex_regex(.inFile$, "[\d]") > 0 \r
 137 \r
 138             # Clean up\r
 139             select Pitch '.inFile$'\r
 140             if .keepIntermediates = 0\r
 141                 Remove\r
 142             endif\r
 143         endif\r
 144     endfor\r
 145 \r
 146     select Table ToneList\r
 147     if .test_word$ <> "" and .test_word$ <> "REUSEPITCH"\r
 148         plus Sound Source\r
 149     endif\r
 150     if .test_word$ <> "" and .test_word$ <> "REUSEPITCH"\r
 151         plus Pitch SourcePitch\r
 152     endif\r
 153     if .keepIntermediates = 0\r
 154         Remove\r
 155     endif\r
 156 \r
 157 endproc\r