Table 2.
Experimental setup with binary intelligibility class for the TORGO database. ‘Num. Samples’ is the number of samples for each speaker. ‘Orig. label’ is the original intelligibility assessment of sentence-level speech audio. The original label of the control group (MC01, MC02, MC03 and MC04) is missing, because they are assumed to be intelligible normal speech.
| Intelligible | Non-intelligible | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Speaker | F03 | MC01 | MC02 | MC03 | MC04 | F01 | M01 | M02 | M04 | M05 |
| Orig. label | a | d/e | d/e | d/e | d/e | c | ||||
| Num. samples | 41 | 186 | 87 | 63 | 157 | 20 | 69 | 50 | 58 | 97 |