Estimation of Phoneme-Specific HMM Topologies for the Automatic Recognition of Dysarthric Speech

Table 5

Recognition accuracies obtained with dysarthric-specific training schemes for the baseline SD ASR systems across all dysarthric speakers.

Speaker	SD Bakis-1	SD Bakis-2	SD Ergodic	Average
Mild scheme
FB	81.22	69.43	76.86	75.84
MH	80.35	72.49	70.31	74.38
BB	77.52	61.47	63.76	67.58
LL	74.11	66.96	65.63	68.90
Moderate scheme
JF	64.00	54.67	52.00	56.89
RL	51.11	47.11	40.00	46.07
RK	39.47	35.53	39.91	38.30
Severe scheme
BK	36.73	30.09	30.53	32.45
BV	56.82	53.64	54.55	55.00
SC	47.27	40.45	41.82	43.18