10 (dB)
Support = (1)
3
Table 1 – Experimental conditions
# Room Position Noise type Noise level
[dB] Rev. time [s] Support [dB]
1 Fontana(a) v1 Babble 57 1.43 -13.6
2 Fontana(a) v1 Ambient 40 1.43 -13.6
3 Fontana(a) v1 Traffic 54 1.43 -13.6
4 Fontana(a) v2 Babble 57 1.47 -15.3
5 Fontana(a) v2 Ambient 40 1.47 -15.3
6 Fontana(a) v2 Traffic 54 1.47 -15.3
7 Fontana(b) v1 Babble 57 0.36 -17.5
8 Fontana(b) v1 Ambient 37 0.36 -17.5
9 Fontana(b) v1 Traffic 55 0.36 -17.5
10 Fontana(b) v2 Babble 57 0.33 -17.8
11 Fontana(b) v2 Ambient 37 0.33 -17.8
12 Fontana(b) v2 Traffic 55 0.33 -17.8
13 Void - Background 28 0.08 -18.1
2.4 Procedure
Before starting the tests, an accelerometer was attached to the subject’s neck, below the glottis. A calibration measurement was performed with the help of a reference microphone, so that it would be possible to estimate the SPL from the accelerometer signal. The head-worn microphone was also attached to the subject, so that the capsule was on the cheek, between the ear and the mouth, at 5 cm from the lips’ edge.
The subjects sat at a chair in the middle of the room. They were instructed to read the
Goldilock’s passage [5], so a listener at 1.6 m could hear it (a dummy head was placed atthat specific position), despite the room acoustics and the noise that was to be presented.
The subjects were asked to start from the beginning of the passage for each condition.
There were a total of 13 conditions, presented in random order for each subject, so that the bias introduced by vocal loading during testing was distributed evenly among the different conditions. For each condition, the reading time was 2 minutes. The start and ending of each presentation was indicated by means of an audio message reproduced through the loudspeakers. Between presentations, subjects had a short break of 30 to 60 seconds approximately. The total experiment time was about 35 minutes on average.
3 Results and analysis
In order to test whether there were significant variations in the different measured vocal doses (Dt, VLI, Dd, De, Dr ) across environments and noises, two sets of ANOVAs (ANalysis Of VAriance) were performed, each one for male and female. The first one tested the hypothesis of equal means between the conditions I) Babble and traffic noise and II) Ambient noise, for each of the environments and vocal doses. The p-values resulting from the ANOVA tests are shown in Table 2. There were significant variations at 5% level in VLI across noises in all the environments. There were also significant variations in Dd across noises in all cases, except in Fontana (b) for female subjects, where nevertheless there were significant variations at 10% level. In the case of De, there were significant variations at 5%
level for male and at 10% level for female. Furthermore there were also significant variations in Dt across noises only for Fontana (a). No significant variations with noises were found for the Dr.
4
The second set of ANOVAs tested the hypothesis of equal means among the two simulated classrooms, for the different doses and noise typologies. The p-values resulting from these ANOVA tests are shown in Table 3, where “no-noise” includes the conditions of background and ambient noise, and “noise” refers to conditions with babble or traffic noise. For male speakers, there were significant variations at the 5% level on all doses, except on Dr, among the different environments, with “no-noise”. Only VLI shows a significant variations at the 10% level in “no-noise” conditions. No significant variations among environments were found in the case of “noise” conditions.
Table 2 – Statistical significance of the variation of measured vocal doses across noise.
Numerical values correspond to p-values from ANOVAs testing the hypothesis of no variation in means across noise, for each combination of vocal dose and environment, and for both
genders.
Gender Male Female
Environment Fontana (a) Fontana (b) Fontana (a) Fontana (b)
Dt 0,01 0,39 0,01 0,23
VLI < 0,01 < 0,01 < 0,01 < 0,01
Dd < 0,01 < 0,01 0,04 0,08
De < 0,01 0,02 0,08 0,09
Dr 0,29 0,24 0,15 0,14
Table 3 – Statistical significance of the variation of measured vocal doses across environments. Numerical values correspond to p-values from ANOVAs testing the hypothesis of no variation in means across environments, for each combination of vocal dose and noise, and for both genders. “No-noise” contains the conditions of background and ambient noise.
Gender Male Female
Noise No-Noise Noise No-Noise Noise
Dt 0,05 0,41 0,13 0,11
VLI 0,02 0,32 0,09 0,97
Dd 0,01 0,50 0,84 0,21
De 0,04 0,33 0,77 0,10
Dr 0,18 0,18 0,15 0,13
The Dt for the different configurations, averaged for 22 subjects, and its standard deviation, are plotted in Figure 2. The Dt accumulated by talkers in “no-noise” conditions was significantly lower than in the other conditions with babble and traffic noise, in good agreement with the low p-values obtained in the ANOVA tests. No significant variations were found between traffic and babble noise.
Figure 3 shows the trend of VLI versus the different configurations, averaged for 11 male and 11 female subjects, and its standard deviation. The trend for both genders is the same. The VLI accumulated by talkers in no-noise conditions was significantly lower than in the other conditions with babble and traffic noise, in good agreement with the low p-values obtained in the ANOVA tests. This effect is related with an increase of the fundamental frequency in conditions with a high level of noise.
The Dr, averaged for 22 subjects, and its standard deviation, are shown in Figure 4. High values of this dose present higher standard deviations. This dose represents the radiated energy from the mouth, and the results show a maximum for both environments in the case of babble noise.
5
60 62 64 66 68 70 72 74
VOID Fontana (a) Tr
Fontana (a) v2 Tr Fontana (a) Ba
Fontana (a) v2 Ba Fontana (a) Bg
Fontana (a) v2 Bg Fontana (b) Tr
Fontana (b) v2 Tr Fontana (b) Ba
Fontana (b) v2 Ba Fontana (b) Bg
Fontana (b) v2 Bg
Dt / s
Figure 2 – Dt as a function of the different configurations, averaged for 22 subjects, and its standard deviation.
6 8 10 12 14 16 18
VOID Fontana (a) Tr
Fontana (a) v2 Tr Fontana (a) Ba
Fontana (a) v2 Ba Fontana (a) Bg
Fontana (a) v2 Bg Fontana (b) Tr
Fontana (b) v2 Tr Fontana (b) Ba
Fontana (b) v2 Ba Fontana (b) Bg
Fontana (b) v2 Bg
VLI / kcycles Female
Male
Figure 3 – VLI versus the different configurations, averaged for 11 male and 11 female subjects, and its standard deviation.
0 10 20 30 40 50 60 70
VOID Fontana (a) Tr
Fontana (a) v2 Tr Fontana (a) Ba
Fontana (a) v2 Ba Fontana (a) Bg
Fontana (a) v2 Bg Fontana (b) Tr
Fontana (b) v2 Tr Fontana (b) Ba
Fontana (b) v2 Ba Fontana (b) Bg
Fontana (b) v2 Bg
Dr / mJ
Figure 4 – Dr as a function of the different configurations, averaged for 22 subjects, and its standard deviation.
6
The values of VLI accumulated by speakers were also considered as a function of the two classrooms, Fontana (a) and (b), in order to investigate the effect of reverberation on this dose. The VLI values were averaged for female and male subjects, across the conditions containing ambient noise. TheVLI scores for both classrooms, along with the standard deviations, are plotted in Figure 5, separately for the results of male and female speakers. In agreement with the p-values obtained in the ANOVA tests, there was a significant main effect of the environment on the VLI accumulated by subjects. In particular, a lower reverberation time resulted in a higher VLI, more for male than for female speakers.
The De values were averaged for female and male subjects, across the conditions with ambient noise. The De scores for both classrooms, along with the standard deviations, are plotted in Figure 6, separately for male and female speakers. There was a significant main effect of the environment on the De accumulated by male subjects, but not for females, in good agreement with the p-values obtained in the ANOVA tests. This dose represent the power dissipated by the vocal fold during the phonation, and it is inversely related to the reverberation time.
6 8 10 12 14 16
Male Female
VLI / kcycles
Fontana (a) Fontana (b)
Figure 5 – VLI accumulated by speakers for the two classrooms.
0 5 10 15 20 25 30 35 40 45
Male Female
De / mJ/cm3
Fontana (a) Fontana (b)
Figure 6 – De accumulated by speakers for the two classrooms.
To investigate possible additional effects of room acoustics on vocal doses, multiple regression analyses of the vocal doses accumulated by the speakers were performed, regarding the values of noise level and the room acoustics parameters (reverberation time and support) as explanatory variables. Table 4 summarizes the results in terms of the
7