1 SpeakGoodChinese tone recognizer
3 The Tone recognizer is an autonomous module. It is implemented
4 completely in Praat scripts. These are run on praat executables
5 either in the background, or bij way of sendpraat commands.
7 The main application calls the scripts with the praat executables.
8 Sound files and resulting texts are written to file
9 (lastExample.wav, lastResult.txt, feedback.txt). There are
10 also some scripts with supporting functions.
12 The feedback advice is stored in the feedback directory and the
13 real advice is generated by ToneRecognition.praat.
14 Currently only positive advice (for tone X you should produce Y).
15 But corrective advice would be possible after adapting
18 The remainder of the files is used for testing the tone models
19 and the recognizer. They generally interface to SGC_ToneProt.praat
22 The crucial tone modelling is done in ToneScript.praat. This script
23 determines syllable structure, duration, and tone models. The actual
24 numbers for synthesizing the tone pitch tracks are read from
25 ToneRules.praat. This latter script encodes the rules in Praat
26 script if-else-endif rules. If you plan to change the tone models,
27 the starting point would be the ToneRules.praat script.
29 Collecting usage and perfomance data, and creating example audio
31 It is possible to record all utterances and write the recognition
32 results to a logFile.txt file. This option is not available from
33 the GUI. Currently, it is switched on by creating a file
34 in the SGC_ToneProt/log directory with the name logPerformance.txt
35 (you can rename an existing stub file DoNotlogPerformance.txt to
36 logPerformance.txt). As long as there exists a file
37 SGC_ToneProt/log/logPerformance.txt, every processed utterance
38 (audio and result) is stored in the directory SGC_ToneProt/log
41 praatcon.exe : Praat background executable for XP
44 SGC_ToneProt.praat : Main script
45 ToneRecognition.praat : The recognizer function
46 ToneScript.praat : Tone generator pinyin -> Tone model
47 ToneRules.praat : "Rules" that determine tone shapes and duration
48 (if you want to alter the tone models, start here)
51 DrawToneContour.praat : Generate and draw a tone model
52 HumToneContour.praat : Generate and hum a tone model
55 CreateFullToneSet.praat : Create a set of all tone shapes for a reiterated syllable
56 SGC_ToneProt.tcl : A TCL/TK interface to test real interactive recognition
57 TestToneRecognition.praat : Run the recognizer on files in a directory
58 ReferenceToneScript.praat : Generate a Pitch object from the pinyin
59 ReplayToneLogFile.praat : Replay the log entries generated by SGC_ToneProt.tcl
62 feedback : Directory with the feedback text
63 feedback.txt : The last feedback text generated
64 lastExample.wav : The last sound used in recognition
65 lastResult.txt : The last recognition result
66 README.txt : This text
67 SGC_protWordList.txt : A wordlist used by SGC_ToneProt.tcl