Table 4 The SVM classification results of PC-GITA database.
From: Phonemes based detection of parkinson’s disease for telehealth applications
Input parameter to SVM | Phoneme | Accuracy (%) | Sensitivity (%) | Specificity (%) |
|---|---|---|---|---|
[SD(Intensity), range(Intensity)] | /a/ | 53.3 | 56.7 | 50.0 |
/e/ | 49.7 | 60.0 | 39.3 | |
/i/ | 56.7 | 64.0 | 49.3 | |
/o/ | 48.3 | 58.0 | 38.7 | |
/u/ | 50.3 | 54.7 | 46.0 | |
/e/ + /o/ + /u/ | 77.3 | 81.3 | 73.3 | |
[Jitt(abs), Jitt(rel), Shim(abs), Shim(rel), SD(pitch), HNR, NHR] | /a/ | 59.9 | 64.4 | 55.3 |
/e/ | 61.5 | 63.1 | 60.0 | |
/i/ | 65.2 | 69.8 | 60.7 | |
/o/ | 61.2 | 63.8 | 58.7 | |
/u/ | 62.5 | 72.5 | 52.7 | |
/e/ + /i/ + /o/ | 70.9 | 74.5 | 67.3 | |
[SD(F1), SD(F2), SD(F3), SD(F4)] | /a/ | 61.3 | 72.7 | 50.0 |
/e/ | 62.0 | 71.3 | 52.7 | |
/i/ | 59.0 | 70.7 | 47.3 | |
/o/ | 57.3 | 72.0 | 42.7 | |
/u/ | 54.3 | 46.7 | 62.0 | |
/a/ + /e/ + /i/ + /o/ + /u/ | 68.0 | 72.7 | 63.3 | |
[VTL(F1), VTL(F2), VTL(F3), VTL(F4)] | /a/ | 69.3 | 70.7 | 68.0 |
/e/ | 65.3 | 64.7 | 66.0 | |
/i/ | 73.0 | 76.0 | 70.0 | |
/o/ | 66.3 | 70.7 | 62.0 | |
/u/ | 63.7 | 67.3 | 60.0 | |
/a/ + /e/ + /i/ + /o/ + /u/ | 84.3 | 84.0 | 84.7 | |
Ten highest-ranked features selected by Relief-F: VTL(F4) of /o/; VTL(F1) of /i/; VTL(F2) of /o/; VTL(F3) of /u/; std(F1) of /o/; std(F2) of /o/; VTL(F1) of /e/; VTL(F1) of /a/; VTL(F2) of /i/; VTL(F2) of /u/ | 71.2 | 70.5 | 72.0 | |
