Evacuation Guidance Assistance System Using Emitting Sound

Abstract

The goal of our research is developing an evacuation guidance system that emits sounds on a set of loudspeakers in a spatial sequence to achieve rapid evacuation in emergency situations. For this goal, we evaluate the auditory recognition properties of emitting sounds and their ability to make evacuees follow the movement of sound stimuli in this chapter. We conduct three experiments to assess the proposed evacuation guidance method. The first and the second ones are done to investigate the recognition properties of the positions and directions of the emitting sound stimuli. The third one is done to evaluate the ability to guide evacuees to exit using the emitting sound in the spatial sequence. In the first experiment, we consider whether the four factors related to the sound-emitting method affect the identification of the emitting sound stimuli. Additionally, we investigate the patterns of emitted sounds that can easily recognize their position and direction. In the third experiment, we consider whether the evacuee can follow the emitting sound on a set of loudspeakers in spatial sequence. Moreover, it is discussed that the proposed guidance system provided a more detailed evacuation route for evacuees.

1. Introduction

In the event of a disaster, such as fire outbreak or earthquake, prompt evacuation to a safe zone is important and essential for ensuring the safety of evacuees. Facility managers are obliged to install fire extinguishing equipment, sufficient emergency exit and guide lights to make evacuee find proper path to the exit by the Japanese Fire Service Law, even if they are not familiar with the layout of the area [1]. Recently, several urban commercial premises have become huge and complex in order to provide efficiency and convenience. Evacuees are not able to intuitively find out evacuation routes in these buildings and structures without proper guidance to exits due to their spatial complexity. Active evacuation guidance systems have been developed to control different light and acoustic stimuli that provide evacuation guidance information to evacuees to construct a safe and secure evacuation environment [2, 3, 4, 5, 6].

Conventional evacuation guidance systems in buildings and structures are not designed to react to changes in situations, such as collapses or other disturbances. Evacuation guidance systems must be able to autonomously determine which evacuation routes have not been damaged by the disaster to achieve rapid evacuation in emergency situations. An autonomous route-detection system, in which several smoke and heat sensors could be placed at key points in the objective area to determine the evacuation route based on the overall condition, was proposed [7]. An evacuation route guidance system that considers evacuees’ current location and building safety using a smart building-sensor network and is able to recommend the best evacuation route for each localized evacuee through their mobile terminals was proposed and evaluated [8]. A method for determining evacuation routes has been proposed that uses location information from mobile devices to determine effective routes [9, 10].

It is crucial to guide the evacuees along the determined evacuation route to effectively use route-detection systems in disasters. However, the evacuation and pathway guide lights in the conventional evacuation guidance were not sufficient to lead the evacuees to the relevant routes adapted to the situation. Therefore, a system is required to guide evacuees flexibly in disasters. Several systems that help evacuees select the route to appropriately exit using light and sound stimuli have been proposed.

Several previous studies have researched active evacuation guidance systems that utilize the precedence effect (Haas effect) to help evacuee realize evacuation direction. It is a psychological feature in hearing acrostic stimuli. When two identical sounds are presented in close succession, the spatial location of the auditory stimulus is dominated by the first arriving sound [11]. Additionally, the implementation of sound equipment (such as loudspeakers and signal processors) and sound-stimuli presentations in evacuation guidance systems have been standardized by the Japan Lighting Manufactures Association (JLMA), such that evacuees can correctly identify evacuation routes [2]. Furthermore, an improved evacuation system utilizing the precedence effect, in which loudspeakers were set beside a wall in a passageway to avoid the disappearance of the precedence effect of an audio signal, was proposed [12, 13].

The evacuation guidance systems using the precedence effect able to lead evacuees to one or two exits predefined as an emergency exit, however, they provide only the direction to the exit and not a detailed evacuation path to the exit using acoustic stimulus. Therefore, the evacuee must discover an evacuation route to the exit using acoustic stimulus even if the guidance using the precedence effect to the exit were provided. If there were several obstacles in the current place, it might not be easy to avoid damaged passageways and a fire outbreak caused by a disaster.

Passengers and crew may quickly lose situational awareness in a smoke-filled cabin of an aircraft. The European Union Aviation Safety Agency (EASA) and Federal Aviation Administration (FAA) regulations stipulate requirements for emergency floor-path illumination in all aircraft to achieve faster evacuation. Thus, guiding pathways to exits makes sense in situations where vision does not work well. However, there are few studies that provide a pathway to exits using emitting acoustic stimuli sequentially. Therefore, in this study, we propose a new active evacuation guidance system using acoustic cues, in which guidance-sound stimuli are sequentially emitted along an evacuation path instead of relying on the precedence effect [6].

This study’s objective is to develop an active evacuation guidance system to direct evacuees along an evacuation route by sequentially emitting sound stimuli. In the first stage of the study, we analyzed participants’ capacity to identify sound stimuli emitted through a set of loudspeakers. We conducted experiments to investigate the recognition properties of the position and direction of the emitting sound, in which four factors, such as the stimulus type and emission-time interval might affect their capacity to identify the stimuli. Additionally, the identification performance of the evacuee for the emitting sequences along the straight and bent lines was considered. Subsequently, we considered whether the evacuee could follow the emitting sound on a set of loudspeakers in sequence. Furthermore, we demonstrated that the proposed guidance system using the sound provided a more detailed evacuation route for evacuees.

The remainder of this paper is organized as follows. Section 2 presents the advantages of the proposed evacuation guidance system that emits sound stimuli sequentially. Section 3 summarizes properties of auditory recognition for the emitting sound stimuli based on our previous research [6]. Section 4 describes the subjects’ ability to follow the sequence of the emitting sound based on experiments and discusses the practicality and feasibility of the proposed evacuation guidance systems. Finally, the usefulness of a sound-based guidance system proposed in this paper is summarized in Section 5.

2. Evacuation guidance systems using emitting sound stimuli

First, an overview and advantages of evacuation guidance systems using emitting sound sources on a set of loudspeakers are described. Conventional guidance systems assume that evacuees can determine their evacuation route based on guide lights in an emergency situation. However, it might be difficult for them to find the relevant or correct route if they are not familiar with the spatial location. Therefore, there are two types of evacuation guidance systems that use acoustic stimuli to indicate the evacuation direction. In this study, we propose a method that uses an emitting sound source on a loudspeaker along routes to guide evacuees. If several loudspeakers are placed in the objective area, the sound sources could be sequentially emitted from one loudspeaker to another. People would be able to recognize them as the stream of sound in the evacuation direction. The recognition of the direction of the sound stream is generalized by the sequence of sound localization for a single-sound source.

In shopping centers, a wide floor is occupied by display cases of goods, which can be an obstacle to evacuation in the event of a disaster. The proposed new evacuation guidance system would be able to provide a pathway to avoid these obstacles because it shows the sequence of the route to exit by emitting sound. This study describes the feasibility and performance of the new evacuation guidance system using the emitting acoustic sound stimuli.

3. Identification of emitting sound patterns in sequences

In this section, as a first step towards developing the guidance system utilizing a sound sequence emitting a sound stimulus to induce people to exit, two experiments to test whether people can identify the sound sequence is described with reference to our previous research [6]. The actual evacuation routes include a variety of patterns, so the two sequence patterns of sounds, straight-line and right-angle patterns were evaluated in the following subsection.

3.1 Identification of emitting sound in straight-line sequences

3.1.1 Factors considered in experiments

First, this study assesses individuals’ capacity to identify a spatial sequence of sound stimuli and examines the factors that influence individuals’ identification performance. Therefore, two experiments were conducted for ten healthy male students (20–21 years old) of Hannan University with no abnormal hearing diagnosed during their annual medical examination. They participated in this experiment without remuneration. Experimenter and all participants provided informed consent before these experiments.

The identification procedure of sound sequence requires the participants to continuously recognize sound localization for sequentially emitting sound through the loudspeakers. The sound localization is a listener’s ability to identify the location of a detected sound in direction and distance. Various factors that cause changes in the pressure and sound wave frequency affect individuals’ sound localization performance. Therefore, it is expedient to consider the influence of emission speed (emission-time interval) and the distance between loudspeakers to assess the subjects’ identification performance. Furthermore, considering the preceding effect, the position of the subject relative to the loudspeaker should be considered a factor affecting the identification of the sound stimulus. Thus, the stimulus type, emission-time interval, distance between loudspeakers, and subject’s position were considered as experimental factors in the first experiment.

Previous studies on emergency-alert sounds reported that stimuli containing a wide range of frequencies are more likely to be recognized than those with a single frequency [12]. Furthermore, swept-sound stimuli from low to high frequencies can be recognized more easily during evacuation guidance procedures. Therefore, in this experiment, the phrase “Here is an emergency exit,” spoken by a female voice in Japanese because the human voice is an acoustic stimulus with many frequencies superimposed on it. It also was used in previous studies on evacuation guidance procedures using the preceding effect [4, 5].

As a result of the above considerations, the two types of acoustic stimulus were set to experimental factor, the female voice and a sound that linearly changed from 500 to 1000 Hz in 1 s (swept-sound). The distance between the loudspeakers was set to two levels, 3 and 5 m, while the emission-time intervals were set to 1 and 0.5 s. The longer the time interval, the faster the sound moved. The subjects were instructed to stand in one of two fixed places—just below a loudspeaker in the center of the grid or between loudspeakers.

3.1.2 Methodology of experiment 1

Our proposed evacuation guidance system help evacuee realize the path to the exit using sequence of acoustic stimuli emitted thought several loudspeakers which were arranged on the ceiling of buildings. In the first experiment loudspeakers were arranged in a 5 × 5 grid 4 m above from floor level in the gymnasium of Hanna University. The experimental environment and the arrangement of the loudspeakers are illustrated in Figures 1 and 2. The numbers rounded with squares in Figure 2 are the index of loudspeakers.

Figure 1.

Experimental environment and the arrangement of the loudspeakers.

Figure 2.

Arrangement of the loudspeakers, subjects’ position, and sequential pattern of sound stimuli in the first experiment.

During the experiment, the subjects stood in one of two possible positions, just below loudspeaker 13 or between loudspeakers 13 and 17. The loudspeakers were a capacitor-type flat speakers with stronger directionality than conventional dynamic speakers. We used a switching device with a small controller (Arduino Uno) that could be controlled by the software to emit the sound stimulus on all 25 loudspeakers in a sequence for specific time intervals. The sound stimulus level was set for each stimulus type (voice or swept-sound), such that the A-weighted noise level was 80 dB at a position 1 m from the loudspeaker. The noise level was measured using an integrated average-type sound-level meter (LA-1441, Ono Sokki Co., Ltd.).

In the first experiment, the sound stimulus (the voice or swept-sound) was emitted sequentially through five loudspeakers. For example, it is the sequence from loudspeakers 1–21 in the order of 1, 6, 11, 16, 21 in a straight line shown by an arrow in Figure 2. All 12 distinct sequence patterns of the sound stimuli (including five row-wise ones (left-to-right or right-to-left), five column-wise ones (front-to-back or back-to-front), and two diagonal ones (front-to-back or back-to-front) are shown by an arrow in Figure 2. Considering that the sequences were emitted in both ascending and descending order, there were totally 24 sequence patterns. For example, the emitting sequence in ascending order was set to five straight line patterns: numbers 1–5, 6–10, 11–15, 16–20, and 21–25.

The subjects were instructed to listen to the sequence at a specific position (as mentioned earlier, position just below loudspeaker 13 or the other) and identify the sequence pattern as quickly as possible. We conducted a trial for each subject to listen to the sound stimuli and answer which patterns were emitted before conducting the first experiment. The first experiment, in which the 24 sequence patterns were emitted randomly in each trial under different conditions (combining the four factors) was then conducted. The sound stimulus continued until the subjects returned their answer. Sixteen trials under each experimental condition combining the four factors were conducted for each subject repeatedly because four two-level experimental factors were considered in this experiment. An experimenter recorded the participants’ response time and the accuracy of their answers (accuracy rate).

3.1.3 Experimental results and discussion for experiment 1

3.1.3.1 Experimental results regarding each factor

Figure 3 shows the mean accuracy rates and response times of identification of emitting sequence regarding four experimental factors. We conducted a three-way analysis of variance (ANOVA) to compare the mean-accuracy rates and response times, considering three within-subject factors: stimulus type, emission-time interval, and the distance between loudspeakers.

Figure 3.

Mean accuracy rates and response times regarding experimental factors. (From Miyoshi [6]). (a) Mean accuracy rates and response times regarding stimulus type. (b) Mean accuracy rates and response times regarding the distance between loudspeakers. (c) accuracy rates and response times regarding interval between sound stimuli. (d) Mean accuracy rates and response times regarding subject’s standing point. p < 0.001: ***, p < 0.01: **, p < 0.05: *.

In the ANOVA results for accuracy rate, significant differences were observed for both stimulus type (p = 0.01) and emission interval (p < 0.01). Similarly, in the ANOVA for the response time, a significant difference was observed in the stimuli type (p < 0.001). Furthermore, an interaction was observed between the emission interval and the distance between the loudspeakers (p < 0.001). Thus, the results of ANOVA for accuracy rate and response time indicate that the sequence of voice stimuli emitted at 1.0 s interval is better than the other sequence.

The influence of four factors on the identification of sequence pattern of sound stimuli was evaluated through the experiment. The overall identification rate was over 80% across 24 sequence patterns. Comparing four factors affecting the accuracy and response time, the significant effects with respect to the type of sound source and the emission-time interval was confirmed. The type of sound source had a particularly strong effect for the results. The accuracy rate of identification was higher, and the response time was shorter when the voice sound emitted as acoustic stimuli than the swept-sound. These results suggest that the use of voice rather than swept-sound as a sound source enables the correct recognition of the direction of guidance.

Aoki conducted the sound localization experiments for middle-aged and elderly subjects using multiple sound sources including voice as acoustic stimuli. As the results it was reported that the incorrect response rate was lowest, and the reaction time was shortest when vocal stimuli were used [13]. Thus, we could easily perceive the vocal phrase and identify its localization. Our experiment task was the identification of the sequence of emitting sound source, and the performance of task become higher for voice stimuli due to the ease of sound location for it.

The differences in the accuracy rates and response times classed according to the different emission-time intervals (1 and 0.5 s) are summarized in Figure 3b). The mean-accuracy rate for both sound stimuli emitted in the interval 1 s was higher than in interval 0.5 s, although there was not a significant difference in response time between both intervals. These results indicate that the stimulus sequence can be identified more easily when the emission-time interval is 1 s.

3.1.3.2 Experimental results with respect to identification of sequence patterns

The accuracy rates and response times of the subjects were compared among sequence patterns, such as row-wise and column-wise sequences. Figure 4a and b illustrate the mean accuracy rates and response times for the row-wise sequence patterns and the left-right direction, respectively. A two-way ANOVA was conducted to detect whether there were statistically significant differences in the mean scores regarding the five row-wise sequence patterns and the left-right directions, considering within-subject factors. It was confirmed that there were the significant differences in the main factor (row-wise sequential pattern) for both the accuracy rate (p < 0.001) and response time (p < 0.001), but there was no significant difference in the main effect regarding the direction and the interaction of two factors.

Figure 4.

Mean accuracy rates and response times for horizontal and vertical patterns in a straight line. (a) Mean accuracy rates regarding row-wise pattern from left and right. (b) Mean response times regarding row-wise pattern on the left and right hand sides. (c) Mean accuracy rates regarding column-wise pattern from left and right. (d) Mean response times regarding column-wise pattern from left and right. (From Miyoshi [6]).

Bonferroni’s multiple comparison (comparison count: 10 times) of the accuracy rate and response time among the 5 stimulus sequences was conducted and its results were shown in Table 1a. The accuracy rate for the row-wise sequence from 11 to 15 was the highest and response time is shortest among the row-wise ones. This result suggests that it is easier to identify the sequence that pass the subject standing point, the loudspeaker 13 and the identification performance became higher as the distance to the sound sequence from subject decreases.

Accuracy rate					Response time
(a) p values for pair-wise comparisons in row-wise pattern
Patterns	1–5	6–10	11–15	16–20	Patterns	1–5	6–10	11–15	16–20
1–5	—	—	—	—	1-5	—	—	—	—
6–10	0.035	—	—	—	6-10	0.002	—	—	—
11–15	<0.001	0.652	—	—	11-15	<0.001	<0.001	—	—
16–20	0.008	1.00	1.00	—	16-20	0.006	1.00	<0.001	—
21–25	1.00	0.001	<0.001	<0.001	21-25	1.00	<0.001	<0.001	<0.001
(b) p values for pair-wise comparisons in column-wise pattern
Patterns	1-21	2-22	3-23	4-24	Patterns	1-21	2-22	3-23	4-24
1–21	—	—	—	—	1-21	—	—	—	—
2–22	0.007	—	—	—	2-22	0.076	—	—	—
3–23	<0.001	1.00	—	—	3—23	<0.001	0.275	—	—
4–24	0.97	0.716	1.00	—	4-24	1.00	0.088	<0.001	—
5—25	1.00	0.057	0.18	1.00	5–25	0.032	<0.001	<0.001	<0.001

Table 1.

Results of Bonferroni’s multiple comparison for accuracy rates and response times in cases of row-wise and column-wise patterns.

In the same way as row-wise patterns, Figure 4c and d illustrate the mean accuracy rates and response times for the column-wise patterns and the direction from front/behind, respectively. The two-way ANOVA also were conducted for the mean scores regarding the sequence patterns and the direction. It was confirmed that the significant differences regarding main effect of sequence pattern were detected in both the accuracy rate (p < 0.001) and response time (p < 0.001) for the column-wise sequence. Additionally, there were the weak significant effect regarding the direction on the accuracy rate (p = 0.078) and the strong significant effect on the response time (p = 0.0068), but the interaction was not detected. Results of Bonferroni’s multiple comparison (comparison count: ten times) of the accuracy rate and response time regarding the column-wise sequences pattern is shown in Table 1b.

The accuracy rate for the sequence from 3 to 23 was the highest and the response time is shortest among the column-wise ones. In the same as results regarding the row-wise sequence, it is easier to identify the sequence that pass the subject standing point, the loudspeaker 13 and the identification performance became higher as the distance to the sound sequence from subject decreases. However, the performances (accuracy rate and response time) for the sequences in the second and fourth line were as well as the third centered line. These results suggest that there may be a range in which people could properly identify the location and the direction of the sound sequence.

3.2 Identification of sound emitting in right-angle sequences

The first experiment investigated the identification performance of the subjects for the straight sequence of the emitted sound. The actual evacuation paths to exit include the straight and right-angle paths. For example, an evacuee evacuates from the inside of a building to the outside by going straight and turning. The identification performance for the emitting sound in right-angle sequences must be evaluated to ensure that the proposed guidance system works effectively in the event of a disaster. In this section, we summarized the second experiment to evaluate the performance of identification for right-angle sequences based on a reference [6].

3.2.1 Methodology of experiment 2

In the second experiment, the sequence of sound stimuli was generated in the same experimental environment with the first experiment as shown in Figure 1. Twenty-five loudspeakers were arranged in a 5 × 5 grid 4 m above from floor level. For this experiment, the distance between the loudspeakers was fixed at 3 m to compare the emission patterns of the straight and right-angle lines. Considering the three experimental factors: sequence shape (two levels of straight and right-angle sequences), stimulus type (voice and swept-sound), and emission-time interval (0.5 and 1 s), the experiment was conducted to assess whether or how these factors affect identification of the sound sequence. The second experiment was conducted under eight experimental conditions, including all combinations of the above three factors. In each trial, sound stimuli were emitted in four right-angle and two straight-line sequences. In addition, the both directionalities of all the sequences, front-to-back one and back-to-front one, were considered, as illustrated by green and bidirectional arrows in Figure 5.

Figure 5.

Arrangement of audio loudspeakers, subject position, and sequential pattern of sound stimuli in the second experiment.

The subjects were instructed to stand at a specific position near by the loudspeaker 23 and listened to and identified the sequence pattern as quickly as possible. The sound sequences were randomly selected from 12 possible sequential patterns and start position of sound sequence was determined randomly. The sound stimulus continued until the subjects returned their answer. An experimenter recorded the participants’ response time and the accuracy of their answers (accuracy rate). The subjects of the second experiment were seven male students (20–21 years old) who had participated in the first experiment.

3.2.2 Experimental results and discussion of experiment 2

Firstly, the mean accuracy rates and response times for each stimulus type and emission-time interval in the second experiment were illustrated in Figure 6. It was confirmed that there were the significant differences in accuracy rates for two factors: stimulus type (p < 0.01), emission-time interval (p < 0.001), and their interaction (p < 0.01), as the result of a two-way ANOVA considering the within-subject factors. Furthermore, there were significant differences in the response times for stimulus type (p < 0.05) and emission-time interval (p < 0.05).

Figure 6.

Mean accuracy rates and response times regarding two experimental factors. (From Miyoshi [6]). (a) Mean accuracy rates and response times regarding stimulus type. (b) Mean accuracy rates and response times regarding the delay of stimuli.

Comparing the identification performance between sound types, the accuracy rate of identification for sound sequence using voice with emission-time interval 0.5 s was less than other cases, but the one for the other conditions was almost 100% because the task was performed completely. The identification of the emitting patterns of voice became difficult when the emitting voice stimulus switched to another speaker in the middle of the phrase. The performance in these emitting conditions became lower than in the other conditions.

Comparing the results for the sequences emitting along the right-angle pattern to the one emitting along the straight-pattern (Figures 3 and 7), the mean-accuracy rates were higher for the emitting pattern at the right angle than in the straight line. In the second experiment, the subjects stood nearby the loudspeaker 23 and identified the sound sequences emitting in front of them. This is because identification performance for the sequence emitting in front of subjects was better than the backward ones as evaluated in the first experiment.

Figure 7.

Mean accuracy rates and response times regarding the spatial proximity to sequence. (From Miyoshi [6]).

Next, the results regarding the sequence patterns are described. The mean accuracy rates and the response times, classed according to the sequence pattern and the spatial proximity to the stimulus from subject standing position were illustrated in Figure 7. The 12 sound sequences were classed into two level in the spatial proximity, “near” and “far” that were indicated with yellow and green lines in Figure 5 respectively.

A two-way ANOVA was used to estimate how the means of the accuracy rates and the response times changed according to the levels of two factors, the sequence patterns and spatial proximity of stimuli from subjects. It was not confirmed that there was neither significant difference in the main factors nor in their interactions. The accuracy rate was almost 100% in all the cases. These results mean that there was no difference in the accuracy rates and response times depending on the sequence pattern, and that the subjects could perceive the sequences of acoustic stimuli emitted in front of them and identify the sequence pattern almost completely in cases where the emission pattern is a straight or right-angle. Therefore, the subjects might be able to follow the acoustic sequence of the emitting stimulus to the emergency exit more quickly even if the evacuation path is more complicate route including straight and right-angle paths. In discussing the practicality of evacuation guidance systems, the results are preferable for realizing a guidance system that emits acoustic stimuli along a predetermined evacuation path.

However, we observed no difference between the sequence shapes, which could be because there was a consistent distance between the loudspeakers. During an actual emergency, the evacuation path may contain a series of short paths to the exit. Therefore, in future, we will verify the subjects’ performance with a smaller distance between the loudspeakers.

4. Ability to guide evacuees by emitting sound along predetermined paths

4.1 Experimental condition of experiment 3

We conducted the third experiment to investigate whether the subjects were able to follow the sound emitted along the evacuation paths. Additionally, we investigated whether it is possible for people to follow complex routes that have a lot of turning points on it. In this experiment, the complexity of the guidance is defined by the number of turning points on itself. As the number of turning points included in a guidance patten increases, it gets more complicated. In case of moving along the more complex path on which subjects turn to the right and left repeatedly in a short time, more accurate and quick sound localization is required to identify the sound stream. Therefore, we defined the complexity of the evacuation path by the number of turning points on it as the difficulty to identify and follow the acoustic stimuli.

In this experiment, the configurations about the locations of loudspeakers and sound sources were the same as those in the previous experiments. Figure 8 shows the experimental configuration and subjects following the sequence. The distance between the loudspeakers and their heights were set to 3 and 4 m, respectively.

Figure 8.

Experimental configuration and a subject following the sound sequence.

The three factors considered important to the performance of the experimental task were the type of sound source, the type of loudspeaker, and the patterns of emitting sound sequences in the third experiment. Two levels, voice and swept-sound, were set with respect to the type of sound source. In the same way with previous experiments, the phrase in female voice, “Here is an emergency exit” was used as voice sound and the acoustic stimuli whose frequency changed from 500 to 1000 Hz continuously was used as the swept-sound.

The two levels, a capacitor-type speaker and a dynamic range one, were set with respect to the type of loudspeaker on which the sound stimuli were emitted. The capacitor-type flat and the conventional dynamic rage speakers have different specifications for directionality, which is the property to focus audio and deliver clear sound precisely where it is needed. In other words, the reduction in sound level through the capacitor-type speaker is smaller even when the reach is farther away because sound spread is smaller than the dynamic one. The capacitor-type one has higher performance than the dynamic range one in the specification of directionality. This factor was designed to evaluate whether the directionality of the loudspeaker affects the sound localization performance for a moving sound source.

The third factor in this experiment is the pattern of emitting sequences. Figure 9 shows the emitting sequence patterns of sound, which are drawn as connections of consecutive column-wise and row-wise line segments. “S” and “G” in Figure 9 indicate the start and goal points of each sound sequence. The sequence patterns are classified into five categories based on the number of turning points in themselves. There were 20 sequences (two sequences in conditions of the number of turning points and start-goal places), as shown in Figure 9. A sequence with one turning point means that the subjects turn for direction once during following it. The third experiment was finally designed under all the combination conditions of three factors (2 levels × 2 levels × 20 patterns) as described above.

Figure 9.

Sound-spatial sequences provided to subjects.

4.2 Experimental procedure of experiment 3

The sound source level of voice and swept-sound were set to 80 dB (A-weighted loudness level) at 1 m from the loudspeaker. The noise level was measured using an integrating average-type sound-level meter (LA-1441, Ono Sokki). The sequences of sound were emitted from the lower left (loudspeaker 21) or lower right (loudspeaker 25) to the upper right (loudspeaker 5) or upper left (loudspeaker 1) which were indicated by circle “G” are shown in Figure 9. The evacuation routes with one to five turning points are also illustrated by the solid and dotted lines in Figure 9, which were formed symmetrically with respect to the diagonal.

The subjects were seven students (six males and one female, 20–21 years old) with no hearing abnormalities during their annual medical examination. The subjects were instructed to follow the sound sequence at walking speed. The sequences of emitting sound were randomly presented to them. In each trial the first point of the sequence was randomly chosen so as not to infer the emitting sequence pattern based on it. The subject stood up at the point under the loudspeaker 23, and identified the sequence of emitting sound and followed it.

The experiment using voice as the sound source was conducted first, and one using swept-sound was conducted two months later. In each experiment the subject performed the trial to identify and follow the sequence of sound that was emitting on two types of loudspeakers, the capacitor-type and the conventional dynamic range speakers. The half of subjects (five subjects) performed the trial using the capacitor-type speakers at first and the one using the conventional dynamic range one after it. The remaining subjects (three subjects) performed the trial in reverse order with respect to the type of loudspeakers.

4.3 Experimental results of experiment 3

4.3.1 Results of the ability to follow the emitting sound

In the third experiment, we recorded the success or failure of following the sequences of the emitting sound and time required to follow the sound from the start and end points. The following three experimental factors were considered in this experiment: the type of sound source (voice and swept-sound), type of loudspeaker (capacitor-flat and dynamic ones), and complexity of the sequence that was defined from one to five by the number of turning points.

The mean values of the time required to follow a sequence and the success rate for each factor are shown in the Figures 10–12. A three-way ANOVA within-subject was used to estimate how the means of the success rates and the required times changed according to the levels of three factors as mentioned above. No significant differences were observed in the success rates for all main factors and their interaction. In the result with respect to the required time, there was no significant difference in the type of sound source (p = 0.457), but there was a significant difference in the type of loudspeaker (p = 0.0415) and the complexity of the emitting pattern (p = 0.0013). Figure 12 shows that, a five-level Bonferroni’s multiple comparison of the complexity of the emitting pattern showed a significant difference.

Figure 10.

Mean values of the time required to follow a sequence and the success rate for sound sources.

Figure 11.

Mean values of the time required to follow a sequence and the success rate for loudspeaker types.

Figure 12.

Five-level Bonferroni’s multiple comparison of the complexity of the emitting patterns.

The average of success rate was 0.970 for all experimental condition, which shows that people were able to identify and follow the sequence of sound in almost all cases. Therefore, there were no significant differences among the sound source, loudspeaker type, and sequence patterns. Even though the success rate of the following is high, the required time is considered as the measure for the difficulty to identify and follow the emitting sound source. In the third experiment, we compared the time required to follow the emitting sound source because the success rate of following the sequences of sound sources was extremely high and no significant difference was observed between the factors.

4.3.2 Experimental results with respect to sound sources

Figure 10 shows that there was the difference in the required following time between types of sound source and it is slightly longer for swept-sounds than voice. However, the analysis of variance showed no significant difference between them in contrast to the results of the first experiment. The task in experimental trial is identifying and following the sequence of emitting sound. The subject is required to repeatedly perform the sound localization for moving sound on the loudspeaker near own current location in the task. The experimental results that the success rates were extremely high illustrate that the subjects were able to perform the continuous sound localization precisely even though the complexity of the guiding routes were relatively high. Therefore, this result suggests that the proposed guidance system using the emitting sound is effective and feasible to lead the predetermined evacuating route even if the voice was used instead of the swept-sound as the acoustic source.

4.3.3 Experimental results with respect to loudspeakers

Figure 11 shows the mean of the required time for following the sequence of the emitting sound on two type loudspeakers, the capacitor flat one and the conventional dynamic range one. The required time is shorter in using the dynamic speaker than the capacitor flat speaker. This result suggests that subjects are able to perform sound localization for a sequence of the sounds emitting on the dynamic range speakers more easily than on capacitor flat one, also were able to follow them. Some subjects commented that they could listen to sounds on the dynamic speaker more clearly than the capacitor one. This tendency is obvious for the sounds emitting on speaker far from their current position. In the task of the third experiment subjects were required to repeatedly perform sound localization for the sound source in spatially wide area. The capacitor flat speaker has stronger directionality than the dynamic range one, then the sound emitting on the dynamic range speaker acoustic stimuli spread more widely and was easier to catch up than on the capacitor one. It is reasonable to assume that the difference of the acoustic property between two types of speakers results the difference of the required time to follow.

4.3.4 Experimental results of sequence patterns

Figure 12 shows a comparison of the required time and the success rates according to the complexity of sequence patterns. There is a little difference in success rates but significant difference (p = 0.0013) in the time required to follow the sequences among the complexity levels by the ANOVA considering the within-subject factors. It was confirmed that the required time increased as the number of turning points increased, i.e., as the complexity of the emitting pattern became higher. The Bonferroni’s multiple comparison (comparison count: 10 times) were conducted for the five complex level of sequences after confirming the significant difference. According to the result, there was no significant difference in the required time between the cases of three and four turning points and between four and five turning points. This result suggests that the difficulty of following the pattern increases as the number of turning points increases.

As the number of turning points in the emitting pattern increases, the success rate slightly decreases, and the required time increases as shown in Figure 12. This result suggests that the evacuation performance decreases as the pattern becomes more complicated in the proposed evacuation system.

5. Conclusion and future works

In this study, the subjects’ ability to identify the location and direction of acoustic spatial sequences and follow it was evaluated through three experiments to discuss the practicality and feasibility of the proposed evacuation guidance systems.

In the first and second experiments, the accuracy rate and response time of subjects for identification the different sequences of sound stimuli were compared among several experimental conditions combining factors: the stimulus the type, emission interval, the distance between loudspeakers, and the sequence patterns. In the first experiment, the accuracy rates improved when the voice stimulus was used and when the emission-time interval was extended. Additionally, it was confirmed that the identification performance becomes better as the distance from the position of subject gets shorter.

In the second experiment, we observed no significant difference in the accuracy rates and response times for different sequence patterns (straight line and right-angle) under the experimental conditions.

In the third experiment the ability of people to follow the sequences of the emitting sound was evaluated based on the success rate and the required time to do so. The three experimental factors were considered, which were the stimulus type, the type of loudspeaker, and complexity of the sequence measured by number of turning points on it. In the third experiment, people took more time but could follow the sequences of emitting sound, which included five turning points. The required time is shorter in using the dynamic speaker than the capacitor flat speaker. This result suggests that subjects can perform sound localization for a sequence of the sounds emitting on dynamic speakers more easily than the capacitor flat one.

The results of this study demonstrate the practicality of an evacuation guidance system using a sound sequence that emits specific sounds on a set of loudspeakers. The factors affecting the performance of subject’s identification of the acoustic stimuli were examined and analyzed, but the level of factors dealt with was limited such as the frequency change region of the swept-sound. Therefore, a more detailed analysis of the degree of influence of each factor is needed in the practical application of the proposed guidance stem for further studies.

Acknowledgments

This research was funded by JSPS Grant-in-Aid for Scientific Research (C) 19K04937. I express my gratitude here.