3 highly damped room, Favrot et al (2010a), Favort (2010b)), and Pelegrín-Garcia et al
3 Results
3.1 The pilot study
The pilot study for this project, Brunskog et al. (2009) showed correlation between the physical characteristics of the rooms and voice power, and with perceived quality, such that the room is perceived good or bad to talk in. It was show that the parameters in the room that primarily affect the voice power are the size of the room and the support for the speaker (or room gain) provided by the room. The results also indicate that the auditory cues may be more important than visual cues in this voice regulation:
the measured changes in the voice power is correlated with the logarithm of the volume (which means a compensation for changes in average noise level outside the room) and not equally well with the cube root of volume (which estimates the mean distance to the audience, that is a visual reference).
3.2 Pre-experiment in the virtual environment
This experiment aimed to investigate the voice used by a teacher to address a group of imaginary students under different simulated acoustics, Pelegrín-Garcia and Brunskog (2009b). Five subjects, aging 23-35 with normal hearing and voice status, talked freely in 5 different simulated acoustic environments during 3 minutes in each of them. The goal was to give a lecture of a familiar topic to a group of 30 students located in front of them. In addition, they had to answer a small questionnaire after speaking in each simulated room. (A more thorough study can be found in Pelegrin-Garcia et al.
(2010a))
Figure 3: Left: Voice power level versus support and regression line. Right: Fundamental frequency variations versus voice support.
The results in Figure 3 show a significant linear dependence (R
2=0.92) between the changes in voice power level used by the speaker and the support provided by the
-16 -15 -14 -13 -12 -11
-4 -2 0 2
Voice Support [dB]
Relative SWL [dB]
-16 -15 -14 -13 -12
-0.04 -0.02 0 0.02 0.04 0.06
Voice Support [dB]
Relative F0 mean
6
3.3 Modified gain
The goal of this experiment was to measure the vocal output when the support was changed, keeping the reverberation time fixed. Thus, the different stimuli did not correspond to actual simulated rooms, but to a single impulse response with different gains. Five teachers talked freely in 10 different simulated acoustic conditions during 3 minutes in each of them. The goal was to give a lecture of a familiar topic to a group of 30 students located in front of them. In addition, they had to answer a small questionnaire after speaking in each simulated room.
The measured variations in voice power level used by subjects are shown in Figure 4.
The trend of the voice power level, indicated by the dashed red line, lays very close to the voice power level measured in the pre-experiment (green line). The slope of the line is in this case -0.58 dB/dB. This indicates that the experiment is fairly repeatable, and that the acoustic environment can systematically change the vocal behavior.
Figure 4: Measured relative voice power level versus support. Dashed red: regression line.
Green: regression line of the pre-experiment.
3.4 Distance to talker
In this subproject the voice used by talkers where analyzed when they had to address a listener at different distances under different acoustic conditions in real rooms. Six talkers, aging 21 to 30, had to talk freely to a listener about a familiar topic during 2 minutes. The listener moved alternately at positions located at 1.5 m, 3 m, 6 m and 12 m away from the talker. The voice of the speaker was recorded from a headworn microphone. This experiment was repeated in four rooms: an anechoic chamber, a reverberation room, a long narrow corridor and a big lecture room.
The measurements show that speakers raise their vocal power when the distance to the listener increases, at a rate of 1.5~2.0 dB per double distance (see Figure 5, left).
The voice power level produced in the anechoic room differed significantly from the other rooms.
-16 -15 -14 -13 -12 -11
-6 -4 -2 0 2 4
Voice Support [dB]
Relative SWL [dB]
7 Figure 5: Left: Variations in Voice Power Level versus distance. Right: Phonation time ratio
versus distance.
Figure 6: Comparison of the model and the measured values. Left: Median voice level vs Support. Right: Median voice level vs. Median noise level.
Furthermore, the subjects expressed their preference about vocal comfort, stating that the least comfortable environments were the anechoic room and the reverberation room. While the analysis of the voice levels cannot account for this preference, other parameters might be better suited. The phonation time ratio (ratio between duration of voiced segments and total duration of running speech) might be appropriated for this purpose, as it can be seen from Figure 5, right. The subjects produce longer vowels in the anechoic room and the reverberation room, compared to the two other rooms, either to overcome the poorer speech intelligibility at the listener location (in the reverberation room) or due to the raised voice levels (in the anechoic room).
3.5 Field study
The field study examines how classroom acoustics interacts with the voices of 14 teachers without voice problems and 13 teachers with voice problems, Pelegrin-Garcia et al. (2010b). The assessment of the voice problems was made with a questionnaire and a laryngological examination. During teaching, the sound pressure level at the teacher’s position was monitored. The teacher’s voice level and the activity noise level were separated using mixed Gaussians. In addition, objective acoustic parameters of Reverberation Time and Room Support were measured in the 30 empty classrooms of the study. An empirical model shows that the measured voice levels (see Figure 6) depend on the activity noise levels and the Support. Teachers with and without voice problems were differently affected by the Support of the classroom. The results thus suggest that teachers with voice problems are more aware of classroom acoustic
1.5 3 6 12
-8 -6 -4 -2 0
Talker to listener distance [m]
Relative VP
Anechoic chamber Lecture hall Corridor Reverberation room
1.5 3 6 12
0.1 0.2 0.3 0.4 0.5
Talker to listener distance [m]
Phonation tim
Anechoic chamber Lecture hall Corridor Reverberation room
-18 -16 -14 -12 -10 -8
657075808590
Support [dB]
Median voice level
8
3.6 Experiments including noise
Another experiment was carried out at DTU in collaboration with the Politecnico di Torino. The goal was to measure vocal doses of speakers under different conditions of room acoustics and noise, Bottalico et al (2010). Vocal doses are a set of measures derived from an estimation of the SPL and the fundamental frequency used by a talker during phonation, Titze et al. (2003). They are measured with an accelerometer attached to the talker’s neck and an Ambulatory Phonation Monitor (APM).
Table 1: Experimental conditions, including noise
# Noise type LN [dB] T30 [s] ST [dB] # Noise type
LN [dB] T30 [s] ST [dB]
C1 Background 28 0.08 -18.1 C8 Traffic 55 0.36 -17.5
C2 Ambient 40 1.43 -13.6 C9 Traffic 55 0.33 -17.8
C3 Ambient 40 1.47 -15.3 C10 Babble 57 1.43 -13.6
C4 Ambient 37 0.36 -17.5 C11 Babble 57 1.47 -15.3
C5 Ambient 37 0.33 -17.8 C12 Babble 57 0.36 -17.5
C6 Traffic 54 1.43 -13.6 C13 Babble 57 0.33 -17.8
C7 Traffic 54 1.47 -15.3