Zoom out Search Issue

ManualsBrandsContents Manualsaudio & home theatreZoom in

121

122

123

124

125

126

127

128

129

130

IEEE SIGNAL PROCESSING MAGAZINE [120] MARCH 2015

, (.,,),maxTFMM0005

-RMSE

,ij

(8)

where (.,,)FMM0 05 corresponds to the F-value computed at

a 0.05 significance level. T 0

,ij

= indicates that metrics i and j

achieved statistically equivalent

f -RMSEs, whereas a T 0

,ij

indicates that metric i is statistically significant worse than .j

EXPERIMENTAL RESULTS

Table 1 presents the results obtained with four intrusive and

four nonintrusive measures on the CI intelligibility database.

Note that results for HASQI, HASPI, PEMO-Q-HI, and SRMR-

HA have been omitted from the table, as they rely on the

impaired listener’s audiogram, which is not readily available

from the CI participants. As can be seen from the table, the

STOI and SRMR-CI measures achieved the highest

sig

t and

lowest

f -RMSE among the tested intrusive and nonintrusive

metrics, respectively. The scatter plots in Figure 1(a) and (b)

depict the subjective versus objective scores obtained for these

two metrics, respectively, along with

their fitted sigmoidal curves.

Table 2, in turn, presents the

results obtained with seven intrusive

and four nonintrusive measures on

the HA nonlinear frequency com-

pression quality database. Note that

the results for SRMR-CI have been

omitted from the table as they rely

on filter bank information from CI

devices. As observed, the PEMO-Q-

HI metric achieved the best

sig

t and

f-RMSE of the intrusive metrics, fol-

lowed closely by the STOI metric

(and the HASQI, in terms of

sig

For the nonintrusive metrics, all

tested measures performed poorly, with ModA achieving some-

what better performance. The scatter plots in Figure 2(a) and (b)

depict the subjective versus objective scores obtained for the

PEMO-Q-HI and ModA metrics, respectively, along with their

fitted sigmoidal curves.

Finally, Table 3 presents the results obtained with seven

intrusive and four nonintrusive metrics on the noisy, reverber-

ant, and enhanced HA quality database. As in Table 2, SRMR-CI

is omitted as it was developed for CI users and not HA. As can

be seen, in the nonenhanced condition, all intrusive measures

achieved similar

sig

t values with PESQ achieving the lowest

f-RMSE, followed closely by STOI. For the enhanced condi-

tion, HASPI achieved the highest ,

sig

t but STOI, PESQ, and

PEMO-Q-HI achieved lower

f-RMSE (over three times lower).

For the nonintrusive metrics, ModA outperformed all others

across both the enhanced and nonenhanced conditions. The

scatter plots in Figure 3(a) and (b) depict the subjective versus

objective scores obtained for the PESQ and ModA metrics,

respectively, along with their fitted sigmoidal curves.

[TABLE 1] PER-CONDITION PERFORMANCE CRITERIA FOR THE CI INTELLIGIBILITY

DATABASE. THE NUMBERS IN BOLD REPRESENT THE BEST ATTAINED PERFORMANCES

(STATISTICALLY INDIFFERENT) AMONG ALL TESTED INTRUSIVE AND NONINTRUSIVE

ALGORITHMS.

ALL

NONENHANCED

(NOISE/REVERB) ENHANCED

METRIC

spear

sig

-RMSE

spear

sig

-RMSE

spear

sig

-RMSE

NCM 0.68 0.74 0.87 9.03 0.96 0.93 0.93 8.41 0.47 0.68 0.77 10.33

STOI 0.81 0.76 0.89 7.05 0.97 0.96 0.97 0.6 0.66 0.69 0.92 3.82

PESQ −0.09 0.01 −0.02 26.85 −0.25 0.4 0.14 26.14 −0.09 0.21

−0.02

23.89

PEMO-Q 0.67 0.53 0.68 15.68 0.72 0.8 0.69 15.67 0.38 0.53

0.44

13.52

P.563 0.05 0.38 0.33 23.59 0.76 0.6 0.78 11.77 −0.79 0 −0.43 25.23

ModA 0.78 0.59 0.78 16.88 0.82 0.76 0.8 13.59 −0.13 −0.17

−0.07

18.42

SRMR 0.49 0.53 0.68 18.41 0.93 0.89 0.92 9.6 −0.35 −0.03

−0.37

23.16

SRMR-CI 0.86 0.77 0.93 5.67 0.98 0.98 0.98 2.06 0.65 0.5 0.88 4.65

0 0.2 0.4 0.6 0.8 1

STOI

Intelligibility (%)

Clean

Noise Only

Reverb Only

Noise−Plus−Reverb

Enhanced

0 0.2 0.4 0.6 0.8 1

SRMR−CI

(a) (b)

Clean

Noise Only

Reverb Only

Noise−Plus−Reverb

Enhanced

[FIG1] Scatterplots of subjective intelligibility versus objective scores for condition-averaged data points obtained from the (a) STOI

and (b) SRMR-CI metrics for the CI intelligibility database.

THE WORLD’S NEWSSTAND

THE WORLD’S NEWSSTAND