Basic study on voice evacuation guidance system using the precedence

effect in a virtual space with wall reverberation Takeru Daimon 1 Department of Electrical Engineering, College of Science and Technology, Nihon University, #1614, 3, Kanda-Surugadai, Chiyoda-ku, Tokyo 101-0062, Japan Ayumu Osumi 2 Department of Electrical Engineering, College of Science and Technology, Nihon University, #1614, 3, Kanda-Surugadai, Chiyoda-ku, Tokyo 101-0062, Japan Youichi Ito 3 Department of Electrical Engineering, College of Science and Technology, Nihon University, #1614, 3, Kanda-Surugadai, Chiyoda-ku, Tokyo 101-0062, Japan

ABSTRACT We have studied a voice evacuation guidance system using the precedence effect, which is a psychoacoustic effect. This system can facilitate rapid evacuation to a safety area in environ- ments with poor visibility, such as those with smoke and dust, because it uses voice announce- ments to guide the evacuation. A basic design for the construction of a sound space suitable for the system and for sound adjustment in the field, such as of speaker position and sound pressure level setting, is required when applying this system to a building. In this research, we propose using virtual reality (VR) technology to support the basic design for constructing this system. If the basic design can be performed in the virtual space using VR technology, then the time required for constructing it can be significantly reduced because there is no need to experiment with it physically. In this study, we constructed a virtual space that reproduced the wall rever- beration of a real space and compared the reflection characteristics of wall between the two spaces. We confirmed that we could obtain the same level of evaluation characteristics for the precedence effect in both the real and virtual spaces.

1. INTRODUCTION

Disaster prevention in large and complex underground buildings is very important. It is presumed that the psychological turmoil of evacuees increases in such buildings during the event of an earth- quake, fire, or other disaster. Additionally, it is difficult to offer evacuation guidance using only evac- uation guide lights when visibility is poor due to smoke. Under these circumstances, evacuees need to evacuate using their hearing only and without any reliance on sight. 1 cste21019@g.nihon-u.ac.jp

2 osumi.ayumu@nihon-u.ac.jp

3 itou.youichi@nihon-u.ac.jp

worm 2022

The focus of this study is to develop an evacuation guidance system that can guide evacuation using sound only. We aim to achieve this by using the precedence effect to give directional sensitivity to the guidance sound [1-6].

The focus of this research is to establish an acoustic design method for implementing this system in various buildings, for example, concert halls, underground malls, etc. Previous studies have suc- cessfully applied this system to straight passages and large interior spaces, confirming that evacuees can be guided in any direction by creating the precedence effect throughout the space [7-8].

On the other hand, when implementing this system in a building, it is necessary to make adjust- ments in the field so that the system can operate optimally under a given set of acoustic characteristics. To address this issue, we seek to achieve the optimal acoustic design for producing the precedence effect by virtually reproducing the sound environment of the building using virtual reality (VR) tech- nology. In previous works, the same reverberation time as in a real space was convolved with the guidance sound in a virtual space to obtain a sound direction sensitivity that is almost the same as the real space [9-10].

As a basic study, we conducted an experiment to evaluate the sound direction sensitivity produced by the precedence effect by simulating it in a virtual space that reproduces the reflection characteris- tics of a wall in a real space.

2. PRINCIPLES OF EVACUATION GUIDANCE SYSTEMS

The precedence effect is a psychoacoustic effect produced when the same sound arrives at the ear from two or more sources located at different positions; the sound is only heard from the direction of the source that reaches the ear first. As shown in Figure 1, to generate the precedence effect, the same sound was played from speaker A and then from speaker B with a delay ranging from several milli- seconds to several tens of milliseconds, so that the listener hears the sound from speaker A only. [11- 16].

worm 2022

Figure 1 : Schematic of the precedence effect.

Figure 2 shows an image of an evacuation guidance system that uses the precedence effect. At this time in the passage, the guidance sounds from the speakers combine through the precedence effect. As a result, evacuees can hear the guidance sound only from the direction of the emergency exit, no matter where they are located.

worm 2022

Figure 2 : Voice evacuation guidance system.

3. EXPERIMENTS FOR EVALUATING THE PRECEDENCE EFFECT IN THE REAL AND VIRTUAL SPACES

3.1. Evaluation experiments in the real space

Figure 3 shows the experiment for evaluating the sound direction sensitivity of the precedence effect in the real space, and Figure 4 shows a schematic view of the experiment. Two speakers, speaker A (preceding sound) and speaker B (succeeding sound), were placed on a line collinear with the participant's ears. Figure 5 shows a block diagram of the setup for playing the guidance sounds. The two speakers played the announcements shown in Table 1. The guidance sounds (60 dB) were played at the position corresponding to the center of the head (3 m from the speaker and 7 cm from the headphones).

The experiment was conducted near the center of a building space measuring 16 m (length) × 13 m (width) × 3 m (height) (with a reverberation time of about 0.9 s and a background noise of about 37 dB), which is shown in Figure 6. The participants wore a head mounted display (HMD) and were presented with the visual information.

Figure 3 : Experimental setup. Figure 4 : Schematic of the experiment.

worm 2022

Figure 5 : Block diagram of the setup for playing the guidance sounds.

(a) Top view. (b) Side view.

Figure 6 : Schematic of the experimental space.

Table 1 : Experimental conditions.

3.2. Evaluation experiments in the virtual space

The virtual space was constructed using Unity, a software tool for developing 3D environments. Figure 7(a) and 7(b) respectively show a schematic and a cross view of the virtual space that was constructed in Unity. The dimensions of this space were identical to those of the real space.

(a) Schematic view. (b) Cross view.

Figure 7 : Virtual space in Unity.

The guidance sound used in the experiment was convoluted using the head-related transfer func- tion (HRTF) of the mannequin head (KEMAR). This HRTF was created and measured by Bill Gard- ner at the MIT Media Lab [17]. In this evaluation experiment, the structures and materials of the real space was considered when reproducing the reflection in the virtual space. This convolved the spatial transfer function with the guidance sound. Unity software (Steam Audio) was also used for signal processing. The guidance sound 𝑃ሺ𝜔ሻ presented to the participants can be expressed by

worm 2022

𝑃ሺ𝜔ሻ= 𝑆ሺ𝜔ሻ× 𝑅ሺ𝜔ሻ× 𝐻ሺ𝜔ሻ , ሺ1ሻ where 𝑆ሺ𝜔ሻ is the sound before signal processing, 𝑅ሺ𝜔ሻ is the spatial transfer function, and 𝐻ሺ𝜔ሻ is the HRTF.

Head movements were detected by using the head tracker function of the HMD. Guidance sounds in the virtual space were played over headphones (SONY, MDR-CD900ST).

3.3. Evaluation of sound direction sensitivity

We conducted the experiment for evaluating the sound direction sensitivity in both the real space and the virtual space. The difference in arrival time between the preceding and succeeding sounds varied from 10 to 100 ms. It is known that the precedence effect becomes significant when the dif- ference in arrival time is more than 1 ms. However, there were a few participants for whom the precedence effect was not perceivable when the difference in arrival time was shorter than 10 ms. Since all participants were able to perceive the precedence effect at 10 ms or more, the experiment was conducted in the range from 10 ms to 100 ms. Participants evaluated the sound direction sensi- tivity using the evaluation values presented in Table 2 and were allowed to move their heads freely during the experiment. A total of 19 individuals (21 to 23 years old) participated in the study.

Table 2 : Evaluation value scale.

4. EVALUATION RESULTS AND DISCUSSION

Figure 8 shows the average evaluation value results for the sound direction sensitivity in the real and virtual spaces. The horizontal axis shows the difference in the time it takes for the sound played by the two speakers to reach the participant. The vertical axis shows the evaluation value of the sound direction sensitivity. In previous work, the evaluation value had to be 4 or above for the system to work effectively. Therefore, a dashed line is inserted in the figure at the position of an evaluated value of 4.

worm 2022

First, we discuss the average evaluation value for the real space. When the difference in arrival time between the preceding and succeeding sounds was within about 70 ms, the evaluation value was above 4. As the difference in arrival time became shorter, the evaluation value became higher. Spe- cifically, when the arrival time was 30 ms or less, the evaluation value remained high and changed gradually. On the other hand, when the difference in arrival time was above 40 ms, the evaluation value dropped sharply. It is presumed that participants could not obtain much sound direction sensi- tivity because the preceding and succeeding sounds were separated. The precedence effect becomes significant when the difference in arrival time is within 50 ms, according to Haas and Bolt et al [15- 16]. The results of this experiment are in good agreement with their findings.

Next, we discuss the average evaluation value for the virtual space. Comparing the evaluation values of the real and virtual spaces, those for the virtual space were generally lower, but the trends of those are same in the two spaces. The difference in arrival time for an evaluation value above 4 was about 55 ms. We also see that the range of the difference in arrival time that produced an evalu- ated value of 4 or higher in the virtual space was about 14 ms shorter than the range in the real space.

We speculate that there are two reasons for this. One was the use of the HRTF of the mannequin. The second is that the reflection characteristics of the wall in the real space are not sufficiently repro- duced in the virtual space.

Figure 8 : Evaluation results for the sound direction sensitivity. 5. CONCLUSION

In this report, we conducted an evaluation experiment on the sound direction sensitivity of the precedence effect in a virtual space in which the structures and materials of a real space are considered when simulating reflected sound. Our results showed that the evaluation values in the virtual space were on average lower than those in the real space due to the error introduced by the HRTF, and to the inability to sufficiently reproduce the reflection characteristics of the wall in the real space. 6. REFERENCES

worm 2022

1. Ito, Y. & Ogushi, K. An intensive system for escape guidance using voice. Journal of the Illumi-

nating Engineering Institute of Japan , 79(9) , 24-29(1995). 2. Ito, Y., Ishiyama, Y., Ishi, H. & Ogushi, K. Study of Escape Guidance with voice Using

Precedence effect in two-dimensional space. Spring Meeting, Acoustical Society of Japan , 859- 860, Tokyo, Japan., March1996. 3. Ito, Y., Sato, H., Ishi, H. & Ogushi, K. A study on recognition time of fusion sound in the Prece-

dence effect. Autumn Meeting, Acoustical Society of Japan , 817-818, Shimane, Japan., Octo- ber1999. 4. Ito, Y. Escape guidance using the sound effect. The Journal of the Acoustical Society of Japan ,

57(10) , 675-680(2001). 5. Daimon, T., Osumi, A. & Ito, Y. Evaluation characteristics of directional judgments under time

constraints in a voice evacuation guidance system using precedence effect. Autumn Meeting, Acoustical Society of Japan , 589-590, Online., September2021. 6. Aoki, S., Maeda, N. & Fujikawa, T. Evacuation system utilizing the precedence effect. Applied

Acoustics , 163 , 107233(2020). 7. Hashimoto, M., Osumi, A. & Ito, Y. A basic study of a voice evacuation guidance system using

precedence effect. Proceedings of INTER-NOISE 2017 , pp. 6929-6934. Hong Kong, China., Au- gust2017. 8. Hashimoto, M., Osumi, A. & Ito, Y. Study on acoustic condition for realizing of a voice evacua-

tion guidance system using precedence effect. The Journal of the Acoustical Society of Japan , 74(9) , 507-510(2018).

Evaluation value Noe RF Ww aA A 0 10 20 30 40 50 60 70 80 90100 Difference in arrival time [ms] -e-Real space -® Virtual space

9. Muto, Y., Osumi, A. & Ito, Y. A basic study of a voice evacuation guidance system in virtual

space using precedence effect. Proceedings of INTER-NOISE 2020 , Seoul, Korea., August2020. 10. Muto, Y., Osumi, A. & Ito, Y. Study of evaluation characteristics of preceding effect using virtual

sound environment. Spring Meeting, Acoustical Society of Japan , 569-570, Online., March2021. 11. Mark, B, Gardner. Historical Background of the Haas and/ or Precedence Effect. The Journal of

the Acoustic Society of America , 43(6) , 1243-1248(1968). 12. Ruth, Y. Litovsky., H. Steven Colburn., William A. Yost. & Sandra J. Guzman. The precedence

effect. The Journal of the Acoustic Society of America , 106(4) , 1633-1654(1999). 13. Rory, W. & Hyunkook L. The effect of interchannel time difference on localization in vertical

stereophony. Journal of the Audio Engineering Society , 63(10) , 767-776(2015). 14. Hans Wallach, Edwin B. Newman and Mark R. Rosenzweig, The Precedence Effect in Sound

Localization, The American Journal of Psychology , 62(3) , 315-336(1949). 15. Haas, H. The Influence of a Single Echo on the Audibility of Speech. Journal of the Audio Engi-

neering Society , 20(2) , 146-159(1972). 16. Bolt, H, R. & Doak, E, P. A Tentative Criterion for the Short-Term Transient Response of Audi-

toriums. The Journal of the Acoustical Society of America , 22 , 507-509(1950). 17. B. Gardner, K. Martin. HRTF measurements of a KEMAR dummy head microphone. MIT Media

Lab Perceptual Computing Technical Report , 280 , (1994).

worm 2022