NAME OF THIS FILE: sim.txt SUSAS Speech Under Simulated and Actual Stress Duke University Robust Speech Processing Laboratory http://www.ee.duke.edu/Research/Speech/ April 15, 1998 (Release Rev. 1.4) October 23, 1997 (Release Rev. 1.1) January 15, 1997 (Release Rev. 1.0) Author: John H.L. Hansen Robust Speech Processing Laboratory Duke University, Dept. of Electrical Engineering Durham, North Carolina 27708-0291 email: jhlh@ee.duke.edu ----------------------------------------------------------------------------- Changes: Release Rev. 1.4 [1] includes label files for all "simulated" stress data, neutral and computer task stress of "actual" data (label files are provided for Scream Machine (/scream) and Free Fall (/freefall) speech, however there may be some phone level shifts due to background noise: they are provided "as is") Release Rev. 1.1 [1] replaced "simulated" data with original 8kHz speech data. Rev1.0 contained "simulated" speech with begin and ending part of the word silence removed. Since some files were cut (stops), we replaced the entire "simulated" portion with the original data. This means that if you obtained label files with Rev1.0, you must re-run with Rev1.1 (or use label files provided). Release Rev. 1.0 Original SUSAS data; includes fixes for missing files from "simulated" portion ----------------------------------------------------------------------------- /susas/sim/ directory contains: Contents: 9 speakers in a quiet environment simulating speech under stress Speakers: --------- general1, general2, general3: Male General USA Accent boston1, boston2, boston3: Male New England/Boston, MA, USA Accent nyc1, nyc2, nyc3: Male New York City, USA Accent Stress Conditions: ------------------ angry 2 tokens of simulated anger clear 2 tokens of clearly enunciated speech cond50 2 tokens of low task stress cond70 2 tokens of high task stress fast 2 tokens of fast speech lombard 2 tokens of speech produced with 85 dBA pink noise in headphones loud 2 tokens of loudly spoken speech neutral 2 tokens of neutral speech question 2 tokens of speech asked as a question slow 2 tokens of slowly uttered speech soft 2 tokens of softly spoken speech train 12 tokens of neutral training speech Special Comments: ----------------- All of these files are in binary short format (16-bit integer) (same as short in C++). All of these files have had excessive leading and trailing silence removed via endpoint detection. Sampled at 8kHz (The data originally donated by Lincoln Laboratory was sampled at 16kHz. A number of errors regarding labels and missing files were identified by us, and appropriate updates made to address these errors included in the present release. These modifications are summarized below. MISSING FILES: -------------- Given the token count for each stress condition and speaker from above, there should be 1,190 tokens per speaker under susas/sim. Unfortunately, a very small number of errors and missing files were present in the data originally collected by Lincoln laboratory. In order to make sure that 2 tokens exist for each stress condition, we did the following: (1) if one token exists for a given stress condition, this file was duplicated to make up the second token: (e.g., break1 did exist, but break2 was missing: therefore break1 was copied to break2) (2) if neither token existed for a given stress condition, then tokens from the neutral directory were used to make up both tokens: (e.g., for speaker general2, under slow conditions, both destin1 and destin2 did not exist: therefore destin1 and destin2 were copied to destin1 and destin2) (3) if a file was originally labeled incorrectly, we used the correct file when possible; otherwise the second stressed speech token is used in it's place: (e.g., the file on1 was actually an original token of the word "oh" (most likely, the speaker said the wrong word): therefore on1 was renamed on1-OH, and on2 was copied to on1) Below is a summary of the 51 files which were either missing, or had the incorrect label when it was originally digitized. The file used to replace the missing entry is also listed [if you are performing stressed speech analysis of SUSAS files, we recommend that you set these files aside. Speaker ORIGINAL REPLACED BY REASON FOR FileLength MODIFICATION ---------------------------------------- general1 1. degree2.g1n actually the word "nav" degree2.g1n was renamed as: nav2 2. fifty2.g1n actually the word "degree" fifty2.g1n was renamed as: degree2 3. nav2.g1n actually the word "wide" nav2.g1n was renamed as: wide2 4. wide2.g1n actually the word "mark" wide2.g1n was renamed as: mark3.g1n-EXTRA 5. 0 fifty2.g1n <--- 7808 fifty1 ORIG. MISSING (after renames) general2 6. 0 destin1.g2s <--- 13952 destin1 ORIG. MISSING 7. 0 destin2.g2s <--- 13440 destin2 ORIG. MISSING 8. 0 fifty1.g2s <--- 7552 fifty1 ORIG. MISSING 9. 0 fifty2.g2s <--- 8064 fifty2 ORIG. MISSING 10. 0 histog1.g2s <--- 12928 histog1 ORIG. MISSING 11. 0 histog2.g2s <--- 12544 histog2 ORIG. MISSING 12. 0 zero1.g2s <--- 20480 zero2 ORIG. MISSING 12b. 0 break7.g2t <--- 8448 break9 ORIG. MISSING general3 13. 0 hot1.g2a <--- 4864 hot1 Orig. token was "break2" (removed and used token1) nyc3 14. 0 thirty2.n3s <--- ORIG. MISSING 15. 0 hello2.n3c5 <--- 11636 hello1 ORIG. MISSING 16. 0 steer2.n3c7 <--- 15520 steer1 ORIG. MISSING 17. 0 gain2.n3n <--- 12800 gain1 ORIG. MISSING 18. 0 no1.n3n <--- 10842 no2 ORIG. MISSING 19. 0 south1.n3n <--- 8192 south2 ORIG. MISSING 20. 0 eight1.n3q <--- 13312 eight2 ORIG. MISSING 21. 0 thirty2.n3s <--- 14848 thirty2 ORIG. MISSING 22. 0 fix1.n3w <--- 10880 fix2 ORIG. MISSING 23. 0 nav2.n3w <--- 10624 nav1 ORIG. MISSING 24. 0 white2.n3w <--- 12928 white1 ORIG. MISSING 25. 0 destin10.n3t <--- 24080 destin11 ORIG. MISSING 26. 0 east2.n3t <--- 12380 east4 ORIG. MISSING nyc2 27. 0 change2.n2c <--- 17024 change1 ORIG. MISSING 28. 0 ten2.n2c <--- 10880 ten1 ORIG. MISSING 29. 0 south2.n2q <--- 15232 south1 ORIG. MISSING 30. 0 thirty2.n2q <--- 11264 thirty1 ORIG. MISSING 31. 0 freeze11.n2t <--- 8576 freeze9 ORIG. MISSING 32. 0 hello5.n2t <--- 7936 hello7 ORIG. MISSING nyc1 33. 0 oh1.n1n <--- 10548 oh2 ORIG. MISSING 34. 0 south2.n1q <--- 7296 south1 ORIG. MISSING 35. 0 three1.n1w <--- 13056 three2 ORIG. MISSING boston1 36. 0 steer1.b1c7 <--- 15744 steer2 ORIG. MISSING 37. 0 break1.b1n <--- 8604 break2 ORIG. MISSING 38. 0 go2.b1s <--- 14336 go1 ORIG. MISSING 39. 0 point7.b1t <--- 11070 point9 ORIG. MISSING 40. 0 six10.b1t <--- 19840 six12 ORIG. MISSING 11264 on1.b1c7 <--- on1-OH renamed;file still exists in dir. 41. 11392 on2.b1c7 <--- 11392 on1 MISSING, since ORIG Wrong Label boston2 42. 0 wide1.b2a <--- 8064 wide2 ORIG. MISSING 43. 0 hello2.b2n <--- 20992 hello1 ORIG. MISSING 44. 0 degree5.b2t <--- 12544 degree7 ORIG. MISSING 45. 0 eighty6.b2t <--- 11264 eighty4 ORIG. MISSING 46. 19102 fifty6.b2t <--- fifty6-FIFTY extra word "an" removed boston3 47. 0 oh2.b3f <--- 4224 oh1 ORIG. MISSING 48. 0 degree2.b3l <--- 9088 degree1 ORIG. MISSING 49. 0 eight1.b3q <--- 8192 eight2 ORIG. MISSING 50. 0 help4.b3t <--- 7040 help6 ORIG. MISSING 30634 nav1.b3s <--- nav1-NAVE.NAV.EXTRA.WORD 51. 20558 nav1.b3s <--- nav1-NAVE extra word "nave" removed