A A A Influential factors on recognition of sound-emitting direction used in an evacuation guidance system Tetsuya Miyoshi 1 dept. of Management Information Hannan University 33-4-5, Amami-Higashi, Matsubara, Osaka 580-8502, Japan. ABSTRACT It is difficult to find appropriate evacuation routes in complicated buildings and underground malls. Therefore, evacuation guidance systems using sound and light have been proposed. An evacuation support system using a sequence of emitted sounds was proposed and evaluated based on its perfor- mance in leading an evacuee to a safe zone. In this system, it was confirmed that several influential factors, such as loudspeaker setting configurations and sound emitting methods, affected the identi- fication performance of the sound sequence. In this paper, we discuss the sound emitting configura- tion of our proposed system to improve the identification performance of the sequence direction. The distance between loud speakers and the emitting time interval are focused on as experimental factors and evaluated based on their effect on the identification performance of the acoustic stimuli. The conditions related to these factors that improve the evacuation support system are discussed. 1. INTRODUCTION Early and quick evacuation is important for minimizing damage to human life during disasters, par- ticularly fires or earthquakes. Evacuation guidelines have been formulated to reduce the damage caused by fires and to shorten the time required to initiate the evacuation and execute it. In most current buildings, it is assumed that evacuees could find their own escape routes based on exit signs and passageway guide lights correctly installed on ceiling according to the given regulations. For faster and safer evacuations, these guidance lights have been improved, and some multimodal active evacuation guidance systems have been proposed that utilize point light sources and sound sources in addition to these guide lights [1, 2, 3]. These are categorized into two types of evacuation guidance system using acoustic stimuli: one based on the preceding sound effect [1, 2], and the other based on a sequence of emitted sounds, which was proposed by us [3]. In this paper, we first summarize the status of multimodal evacuation guidance, including point light sources and sound sources such as active guidance systems. These systems aim for faster and more reliable guidance. Next, we describe the evacuation guidance system using sequences of emitted sounds on a set of loudspeakers, which we have proposed in the last conference [3]. In our proposed system, the identification performance of the acoustic sequence of the emitted sound must be im- proved to make it effective and feasible. We conducted an experiment to investigate the characteris- tics of identification of the acoustic sequence with respect to the sound emitting configuration, time interval of the emitted sound, distance between loudspeakers, and starting point of the sequence. We report the experimental results later in this paper and discuss their effect on the performance. 1 miyoshi@hannan-u.ac.jp worm 2022 2. MULTIMODAL EVACUATION GUIDANCE SYSTEM This section summarizes the status of current evacuation guidance systems using point light sources and acoustic stimuli. Additionally, smart guidance systems collecting information on mobile phones about available passages and guiding evacuees to a safe area are reported. 2.1. Need for active evacuation guidance system It is recognized that quick evacuation can reduce the death toll of a disaster. As stated by the law in Japan [4][5], an installation of evacuation exit guide lights is required for fast evacuation in the event of a fire or other disaster in large buildings such as commercial shopping malls and underground malls with a high number of visitors. The passageway guide lights need to be installed according to the distance to an evacuation exit. As described above, the evacuation guidance system regulated by the national law in Japan assumes that evacuees could search for passages to the exit and determine a relevant evacuation route by using the exit and passageway guide lights. Recently, buildings have become taller and extended deeper underground to effectively utilize space in urban areas; in such cases, it may be difficult to find an appropriate evacuation route during a disaster due to the spatial complexity and variety. During a disaster, normal evacuation routes might be inaccessible due to the many obstacles created by the disaster. Moreover, ordinary guides only show the direction to evacuation exits, but not the routes to avoid them. These matters cause difficulty in finding appropriate routes in complex buildings, even if the exit signs and passageway guide lights are installed according to the guidance. Considering these circumstances regarding evacuation support, it is recognized that evacuation guidance systems must actively provide evacuees more information about the evacuation. Therefore, awareness of exit lights is being improved by increasing its luminance and location awareness by making the lights multimodal, for example, by using flashing lights and emitting guiding sounds. Furthermore, considering the need to provide useful information on evacuation, evacuation routes are presented using point light sources and sound sources. Additionally, evacuation guidance systems using mobile devices have been proposed lately [6]. 2.2. Guidance using light stimulus To provide active evacuation guidance, it is necessary to determine an appropriate evacuation route considering the collapse and fire conditions that occur during a disaster and to provide efficient and reliable guidance along that route. This section focuses on some evacuation guidance methods and summarizes the status of evacuation guidance systems using point light sources and sound sources. The effectiveness of an evacuation guidance system in which a point light source runs along the evacuation route to indicate the evacuation route has been verified. Attachment conditions such as size, luminance, spacing, and flashing running speed of the light source when implemented are spec- ified [7]. It is also effective in evacuation guidance for the hearing impaired, and its implementation is recommended from the perspective of universal design [8]. In aircraft evacuation guidance, the embedding of flashing running lights in the floor is mandatory to support rapid evacuation after an accident [9]. worm 2022 2.3. Guidance using sound stimulus Sound localization, which identifies the sound source position and direction from the source, is pro- cessed based on the arrival time and sound pressure differences of the sound information heard by the left and right ears. Low-frequency sounds are less likely to cause binaural time difference (interau- ral time difference) ambiguity due to their longer wavelength, whereas high-frequency sounds are less likely to diffract, resulting in a binaural sound pressure difference (ILD, interaural level differ- ence). It is well-known that in basic human auditory information processing, sound source localiza- tion is based on these frequency-based perceptual characteristics of the sound source [10]. It has been reported that a guidance system utilizing auditory characteristics related to source localization can effectively guide evacuees, since evacuees can recognize the direction of evacuation quickly [11]. It was also illustrated that the evacuation guidance system could lead evacuees to the exit even if the distance to the exit was high, using several loudspeakers installed along the route over a wide area [12]. The Haas effect is a well-known human auditory characteristic, in which a sound source perceived slightly earlier is perceived as dominant. A proactive evacuation guidance method that applies the sound effect has been proposed [1,2,12]. In this evacuation system, loudspeakers are placed in the four corners of the target area, and the loudspeaker in the direction to be guided emits the sound a bit earlier than the other loudspeakers, followed by the other loudspeakers. This induces a precedence effect on people; they can recognize the sound emitted from the loudspeaker in the direction of in- ducement as the dominant sound. The evacuation support system could guide evacuees to the exit effectively, and its effectiveness in virtual spaces has been verified [12]. Unlike evacuation guidance that applies to the preceding sound effect, an evacuation guidance method that emits sound sources on loudspeakers along the evacuation paths has been proposed [3]. In our recent study, the identification of the emitted sound on loudspeakers and the possibility of following the emitted sound were evaluated, and it was reported that a high percentage of emitted sound recognition was possible. In this research, experiments in which several subjects followed se- quences of emitted sounds in several experimental conditions were conducted, and the efficiency of guidance and the complexity of the guidance pattern were quantitatively evaluated. The experimental results reported that a success rate of following the stream of acoustic stimuli reduced as the com- plexity of the sequence increased , but it was possible to follow it with a relatively high rate. 3. EVALUATION OF RECOGNITION FOR DIRECTION OF EMITTED SOUND In this section, we evaluate our experimental method and the results to identify the appropriate con- figuration for facilitating easy recognition of the guidance direction. 3.1. Configuration of Emitted Sound and Background Noise In our proposed evacuation guidance system, it is assumed that many loudspeakers are placed on the ceiling of a target area, and sound sequences of sound stimuli emitted through ones along the pathway to an exit are provided to the evacuee. The identification performance of the location and direction of the sound sequence generated in the system was discussed with respect to experimental factors such as types of sound sources, distance between loudspeakers, and interval time between emitted sound in the previous paper [3]. Additionally, it was investigated whether the sequence patterns, straight or bending patterns, affect the identification performance. In this paper, the proper configuration for providing the sound sequences is discussed. For this aim, we deal with a straight sequence of the emitted sound as a guidance route. The loudspeakers were arranged in a straight line on the assumption that they were set on ceiling, as shown in Figures worm 2022 1 and 2. Twenty-five loudspeakers were placed at intervals of 0.5 m in a 12 m straight line. Each number surrounded by a square indicates the speaker number in Figure 1. The loudspeakers were installed at a height of 3 m above floor as shown in Figure 2. Experiments were conducted in conditions where the background noise level was 33.0 dB at the A-weighted time-averaged sound pressure level, and in the noisy condition with white noise present. In the latter case, the noise generators were placed at 2 m behind loudspeaker 5 and 21, respectively, and the noise level was set to 60 dB (the A-weighted time-averaged sound pressure level) at the subject's position. The sound sequence is used as a guidance source to an exit in the proposed evacuation guidance system. The phrase "Here is the emergency exit" is vocalized with female voice using the text-to- speech software, and it was emitted through each loudspeaker sequentially with a predetermined time interval. The sound source length was adjusted to one second. Prior to the experiment, the sound level of the source was set to 60 dB (the A-weighted time-averaged sound pressure level) at 1 m from the loudspeaker. The noise levels of the emitting source and noise were measured using an integrally averaged common sound level meter (LA-1441, Ono Sokki). Loudspeakers for guidance Subject 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 2.0m worm 2022 0.5m Loudspeaker for noise 12m Figure 1. Loudspeaker placement in the experiment Figure 2. Experimental configuration of emitted sounds 3.2. Effect of experimental factors on the identification performance We investigate the appropriate sound configuration in the evacuation guidance system through this experiment. Although many factors might be considered as influential factors for the identification, we focused on two of them, which were considered as stronger factors. The first one is the distance between loudspeakers, and the second is the time interval in the sound sequence. We discuss their impact on identification performance. Two levels of loudspeaker spacing (0.5 m and 1.0 m) and three levels of time intervals (0.2 s, 0.5 s, and 1.0 s) were used for the two factors. As the loudspeakers were placed at 0.5 m intervals, the control program of emitted sound was set up so that the sound sources were emitted on every loudspeaker sequentially, in case the loudspeaker distance was 0.5 m, and on every other loudspeaker in case the distance was 1 m. If the start position of the sound se- quence was fixed, the emitting direction could be inferred from the start position, so the emitting start position when presenting the sound source was changed as shown in Figure 3. The start points of the sequence were set to 5 and are indicated by "S" in Figure 3. The emitting order was an ascending and descending one, and the starting points were determined to be symmetrical between the two. As a result, the number of sequence patterns was ten. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 loudspeaker arrangement ascending sequence pattern S P1 S P2 S P3 S P4 S P5 descending sequence pattern S P1 S P2 S P3 S P4 S P5 Figure 3. Sequence patters of emitted sound 3.3. Instruction to subjects and measurements in experiment The eight subjects were healthy male students (20–21 years old) of Hannan University with no ab- normal hearing diagnosed during their annual medical examination. They participated in this experi- ment without remuneration. The experimenter and all participants provided informed consent before these experiments. worm 2022 For each of the six conditions (three levels for both factors), the subjects were randomly presented with emitted sounds in a total of 60 different sequence patterns, with five patterns varying the starting position in ascending or descending order of loudspeaker number, as shown in Figure 3, and asked to respond as quickly as possible in the emitting direction. Subjects' response times and correct or in- correct responses were recorded for each trial. 3.4. Results on accuracy rate and response time related to experimental factors (1) Results related to the emitting interval and distance between the loudspeakers The experimental results for background noise (33 dB) without added environmental noise are shown. The mean accuracy rate of response time for the loudspeaker intervals and the emitting time intervals as experimental factors are shown in Figures 4 and 5, respectively. As shown in Figures 4(a) and 5(a), the accuracy rates of responses were almost 100%. A within-subjects two-way analysis of variance was conducted on the accuracy rate, with loudspeaker interval and emitting time interval as factors, and no significant differences were identified for both the loudspeaker interval ( p = 0.351) and emit- ting time interval ( p = 1.00). Identification of the direction of the sequence seemed to be a simpler and easier task than the ones in the previous research because the location and direction of flow of the sequences passing at the subject position are easy to identify. This is the reason why all of them could be correctly answered. The mean response time by loudspeaker interval and the results of ANOVA are shown in Figure 4(b), indicating that the longer the speaker interval, the shorter the response time. The mean response time by emitting time interval is shown in Figure 5(b), in which the response time increased as the emitting time interval increased. A within-subjects two-way ANOVA on response time with speaker interval and emitting time interval as factors confirmed a significant main effect for both speaker interval ( p <0.001) and emitting time interval ( p <0.001). No interaction ( p =0.232) was identified between the two factors of loudspeaker interval and emitting time interval. Bonferroni's multiple comparisons (number of comparisons: three) for emitting time intervals also confirmed significant differences for all combinations of 0.2s and 0.5s ( p <0.0001), 0.2s and 1.0s ( p <0.0001), and 0.5s and 1.0s ( p <0.0001). Furthermore, the mean of the response time of each subject related to two factors is shown in Figures 6 and 7, respectively. These figures show that the same relationship between emitting time interval and response time was confirmed in all subjects' responses. Considering these results, it can be confirmed that the longer the emitting distance of the acoustic stimulus per unit time, i.e., the faster the speed of movement of the emitted sound, the shorter the time required to recognize the emitting direction. These results suggest that emitting direction recog- nition requires emitting sounds at a certain emitting distance, and therefore, the time required for emitting direction recognition decreases as the emitting speed increases, i.e., when the loudspeaker interval during emitting is long or the emitting time interval is short. worm 2022 8.00 0.00 0.20 0.40 0.60 0.80 1.00 response time(s) 6.00 *** accuracy rate 4.00 2.00 0.00 0.5m 1.0m 0.5m 1.0m distance between speakers distance between speakers p < 0.001: ***, p < 0.01: **, p < 0.05: * Figure 4. Mean accuracy rate and response time for the distance between loudspeakers and significant difference between 2 levels of the distance. 0.00 0.20 0.40 0.60 0.80 1.00 8.00 *** *** *** response time(s) 6.00 accuracy rate 4.00 2.00 0.00 0.2s 0.5s 1.0s 0.2s 0.5s 1.0s time interval time interval p < 0.001: ***, p < 0.01: **, p < 0.05: * Figure 5. Mean accuracy rate and response time for the time interval of emitted sounds and significant difference between 3 levels of them. 0.0 2.0 4.0 6.0 0.0 2.0 4.0 6.0 6.0 6.0 respons time(s) response time(s) response time(s) response time(s) 4.0 4.0 2.0 2.0 0.0 0.0 0.5m1.0m 0.5m 1.0m 0.5m 1.0m 0.5m 1.0m distance distance distance distance worm 2022 6.0 6.0 6.0 6.0 response time(s) response time(s) response time(s) response time(s) 4.0 4.0 4.0 4.0 2.0 2.0 2.0 2.0 0.0 0.0 0.0 0.0 0.5m 1.0m 0.5m 1.0m 0.5m 1.0m 0.5m 1.0m distance distance distance distance Figure 6. Response times of all subjects for the distance between loudspeakers 0.0 2.0 4.0 6.0 8.0 0.0 2.0 4.0 6.0 8.0 0.0 2.0 4.0 6.0 8.0 0.0 2.0 4.0 6.0 8.0 response time(s) response time(s) response time(s) response time(s) 0.2s 0.5s 1.0s 0.2s 0.5s 1.0s 0.2s 0.5s 1.0s 0.2s 0.5s 1.0s time interval time interval time interval time interval 0.0 2.0 4.0 6.0 8.0 0.0 2.0 4.0 6.0 8.0 0.0 2.0 4.0 6.0 8.0 0.0 2.0 4.0 6.0 8.0 response time(s) response time(s) response time(s) response time(s) 0.2s 0.5s 1.0s 0.2s 0.5s 1.0s 0.2s 0.5s 1.0s 0.2s 0.5s 1.0s time interval time interval time interval time interval Figure 7. Response times of all subjects for the time interval of emitted sound (2) Result related to emitted patterns Figure 8 shows the mean accuracy rate and the response time for each sequence patterns. In sequences P1 and P2, the acoustic stimuli were approaching a subject in both ascending and descending patterns, whereas in sequences P3 and P4, one was going away from a subject. P5 is the sequence in which the acoustic stimuli are emitted at the end of the placed loudspeaker and then approach a subject from the other side. Because the task of identifying one of the two directions was simple, it was performed, almost completely, regardless of the sequence pattern. However, the differences in response time were observed depending on the pattern. A within-subjects one-way ANOVA on response time with sequence patterns confirmed a signifi- cant main effect for sequence patterns ( p <0.001). Bonferroni's multiple comparisons (number of comparisons: ten) for the sequence patterns also confirmed significant differences between P2, P3, and the others. The differences are shown in Figure 8(b). The results show that the response time for sequences P2 and P3 were short whereas those for the others were long. The results suggest that the subjects need to listen to a sound source moving over a certain distance to recognize the direction of the source, but they can recognize it within a short distance in the case of a sequence of patterns that are close to the subjects. This should be discussed in comparison with the results obtained under conditions in which the listening range is narrowed by noise. worm 2022 *** 0.00 0.20 0.40 0.60 0.80 1.00 0.00 1.00 2.00 3.00 4.00 5.00 *** *** response time(s) accuracy rate *** *** P1 P2 P3 P4 P5 P1 P2 P3 P4 P5 emitting pattern emitting pattern p < 0.001: ***, p < 0.01: **, p < 0.05: * Figure 8. Mean of accuracy rate and response time for emitting sequence patterns. 4. CONCLUSIONS This paper first described the necessity of providing multimodal information, in addition to the cur- rent evacuation guidance methods, using emergency exit lights and guide lights to help reduce human casualties during disasters. The paper then summarized the evacuation guidance method using sound and flashing lights as multimodal information. Next, the method of guiding evacuation by sound sequences was introduced, and to make it effec- tive and feasible, the performance of identifying the acoustic sequence of the emitted sound was investigated and many matters were clarified. In the case of the sound sequence, a certain distance of scanning is required for recognition, and it is shown that the recognition time can be reduced by increasing the scanning speed. For the recogni- tion performance of the scanning pattern, the time to recognition was shortened in the case of a sound coming toward the listener. Future work should analyze the scanning distance required for the recognition of scanning direction of scanning sound sources in detail. 5. ACKNOWLEDGEMENTS This research was funded by JSPS Grant-in-Aid for Scientific Research (C) 19K04937. I express my gratitude here. 6. REFERENCES 1. Muto, Y., Osumi, A., Ito, Y., A basic study of a voice evacuation guidance system in virtual space using precedence effect, Proc. of InterNoise20, 1690-1697, (2020). 2. Aoki S., Maeda N. and Fujikawa T. Evacuation system utilizing the precedence effect. Applied Acoustics, 163(June), DOI: 10.1016/j.apacoust.2020.107233 2020. worm 2022 3. Miyoshi, T., Evaluation of recognition performance of guidance direction presenting by sequen- tial emitting sound in evacuation guidance system, Proceedings of InterNoise21, pp. 5335-5345, DOI: https://doi.org/10.3397/IN-2021-3052. 4. Fire and Disaster Management Bureau, Ministry of Internal Affairs and Communications, "Fire Service Law Enforcement Order Article 26" ， https://elaws.e-gov.go.jp/docu- ment?lawid=336CO0000000037 ( in Japanese, accessed 1 May, 202). 5. Fire and Disaster Management Bureau, Ministry of Internal Affairs and Communications, Stand- ards for Guide Lights and Signs, https://www.fdma.go.jp/laws/kokuji/assets/h11_kokuzi2.pdf (in Japanese, accessed :1 May 2021). 6. Hara T., Kasabe M., Kasahara M. On path selection considering length and reliability for auto- matic evacuation guiding based on interactions between evacuees and their mobile nodes, IE- ICE Technical Report, 116(484), 517-522, 2017. 7. Ronchi, E., Nilsson, D., Kojić, S. et al. A Virtual Reality Experiment on Flashing Lights at Emer- gency Exit Portals for Road Tunnel Evacuation. Fire Technol 52, 623–647 (2016). https://doi.org/10.1007/s10694-015-0462-5 8. Fire and Disaster Management Bureau, Ministry of Internal Affairs and Communications, Guide- lines for the Installation of Optical Alarm Devices, https://www.fdma.go.jp/singi_kento/kento/items/kento160_23_sanko2-2.pdf (in Japanese, ac- cessed 1 May 2021). 9. Paskoff, L. N., et al., Wayfinding Technology and Its Application to Transport Category Passen- ger Airplanes, Report No. DOT/FAA/AM-15/14, Federal Aviation Administration, (2015), https://www.faa.gov/data_research/research/med_humanfacs/oamtechreports/2010s/me- dia/201514.pdf , (Accessed: 1 May 2021). 10. Rayleigh, L., On our perception of sound direction, Phil Mag., 13(6), pp.214-232, (1907). 11. Withington, D. J., Localizable alarms, in N. A. Stanton, J. Edworthy (eds), Human Factors in Auditory Warnings, pp.33-40, (1999). 12. Plocher, T., Jin, Z. J., Chan, F. Y. D. (2011). Using Sound Patterns to Enhance Directional Sound for Emergency Route Guidance. In: Jacko, J.A. (eds) Human-Computer Interaction. Towards Mobile and Intelligent Interaction Environments. (2011). vol 6763. Springer, Berlin. https://doi.org/10.1007/978-3-642-21616-9_37 . 13. Hashimoto, M., Osumi, A., Ito, Y., A study of a voice evacuation guidance system using a prec- edence effect, Journal of the Acoustical Society of America, 140(4), 3275-3275,(2016). DOI: 10.1121/1.4970404 worm 2022 Previous Paper 513 of 769 Next