Table 5.
Feature set
|
Results
|
||||||
---|---|---|---|---|---|---|---|
Estimation
|
Decoding
|
||||||
Feature type | Subject(s) | Phoneme error rate (%) | Posteriogram accuracy (%) | Confusion accuracy (%) | Phoneme error rate (%) | Posteriogram accuracy (%) | Confusion accuracy (%) |
Chance | − | 97.12 ± 5.34 | 7.06 ± 9.77 | 2.63 ± 16.01 | 97.12 ± 5.34 | 7.06 ± 9.77 | 2.63 ± 16.01 |
MFCC | − | 208.16 ± 67.45 | 40.45 ± 10.76 | 34.51 ± 16.15 | 60.60 ± 25.16 | 41.88 ± 17.30 | 36.30 ± 10.86 |
Single-trial HGS | A | 188.73 ± 90.28 | 11.74 ± 7.98 | 7.67 ± 10.03 | 90.84 ± 25.00 | 11.79 ± 10.88 | 11.49 ± 7.94 |
B | 176.91 ± 114.31 | 7.72 ± 6.02 | 4.79 ± 4.96 | 92.31 ± 26.43 | 10.23 ± 11.06 | 9.63 ± 5.52 | |
C | 173.83 ± 92.48 | 9.00 ± 6.76 | 4.80 ± 7.22 | 87.67 ± 17.36 | 10.70 ± 11.58 | 8.16 ± 6.23 | |
Single-trial HGW | A | 169.68 ± 84.89 | 15.62 ± 9.21 | 12.76 ± 10.12 | 86.34 ± 22.07 | 15.06 ± 11.73 | 14.89 ± 9.01 |
B | 151.70 ± 78.01 | 12.69 ± 8.18 | 9.58 ± 5.89 | 90.63 ± 26.28 | 12.71 ± 11.49 | 11.80 ± 5.96 | |
C | 152.48 ± 75.98 | 12.93 ± 8.77 | 8.09 ± 8.22 | 85.71 ± 17.30 | 12.62 ± 12.10 | 9.84 ± 6.40 | |
Averaged HGS | A | 182.53 ± 89.81 | 14.68 ± 8.87 | 10.81 ± 10.80 | 86.37 ± 20.83 | 14.03 ± 11.72 | 14.07 ± 9.28 |
B | 182.80 ± 99.75 | 11.95 ± 7.82 | 8.21 ± 6.92 | 87.13 ± 18.01 | 13.51 ± 11.30 | 12.59 ± 7.26 | |
C | 181.75 ± 78.74 | 12.22 ± 8.01 | 7.08 ± 9.01 | 85.36 ± 13.67 | 12.41 ± 11.45 | 10.01 ± 6.56 | |
All | 180.87 ± 84.47 | 22.34 ± 9.94 | 18.14 ± 10.47 | 76.68 ± 15.82 | 23.41 ± 13.85 | 21.39 ± 9.12 | |
Averaged HGW | A | 158.00 ± 69.86 | 18.93 ± 10.39 | 16.49 ± 10.99 | 81.39 ± 18.85 | 18.79 ± 12.57 | 18.49 ± 10.30 |
B | 149.66 ± 65.27 | 17.35 ± 9.46 | 13.80 ± 7.30 | 82.84 ± 21.02 | 17.64 ± 13.31 | 15.30 ± 6.99 | |
C | 157.35 ± 73.80 | 16.42 ± 9.84 | 11.20 ± 8.79 | 83.65 ± 19.18 | 15.36 ± 13.02 | 12.27 ± 7.99 | |
All | 142.96 ± 61.31 | 28.36 ± 10.47 | 24.02 ± 10.99 | 70.47 ± 16.77 | 29.26 ± 14.88 | 25.03 ± 10.69 |