Speech Recognition

1
Speech Recognition

• Progress Report

2
Database Conclusions

• The RES database is made up of wave files and
  phoneme files (TIMIT)
• Phoneme files can be created from any sound file
  with PRAAT
• The RES will access these files through an
  ordered file system

3
Jobs

• Figure out the ordered system based on TIMIT
• Write a PRAAT script to convert AIFF files from
  the IFA corpus to the TIMIT format
• Decide what to put into our RES system

4
TIMIT format vs PRAAT
Praat
Timit
xmax 6.1032500000000001
intervals size 70 intervals 1
xmin 0 xmax
0.50773239999999997 text
"__-A1SK1" intervals 2
xmin 0.50773239999999997 xmax
0.5604595 text "d__F20N1FR1C1SO1"
0 1994 h 1994 2760 hh 2760 3400 iy 3400 4469
hv 4469 5723 eh 5723 7001 z 7001 7591 n
5
Deadlines

• First modules available by next week