Enhanced 3D acoustic scene analysis based on sound arrival direction for automatic airport noise monitoring. Keishi Sakoda 1 RION CO. Ltd. 3-20-41 Higashimotomachi, Kokubunji, Tokyo 185-8533. Japan Ichiro Yamada 2 RION CO. Ltd. 3-20-41 Higashimotomachi, Kokubunji, Tokyo 185-8533. Japan

ABSTRACT To deepen our understanding of the spatial and temporal distribution of various sound sources of environmental noise, which is observed by unattended aircraft noise monitors, we have been developing a method of acoustic scene analysis based on information on the 3D sound arrival direction. In the congress of INTERNOISE 2021 last year, we reported the basic idea of this method and some examples of the analysis. The method of acoustic scene analysis, named static mode and dynamic mode, is based on the information of the direction of arrival and sound pressure level of the sound in three dimensions from time to time. In this report, we describe the current status of efforts to improve the functionality of these two modes. In order to describe the results of the trial of the video availability for the static mode. The purpose is to create a document that will facilitate a deeper understanding of the sound environment. In order to improve the versatility of the Dynamic Mode Acoustic Scene Analysis method, we decided to improve the method's versatility. Because the travel paths are limited to those that satisfy the specified preconditions, there were restrictions on the sound sources moving over the applicable airspace. Therefore, by adding a sensor that observes the direction of sound arrival, the preconditions can be excluded. In this paper, we report on the study and experimental status.

1. INTRODUCTION

In a previous paper 1) , we discussed a method of evaluating three-dimensional (3D) acoustics scenes by utilizing the information of instantaneous 3D sound arrival directions and sound pressure levels observed at a site of unattended aircraft noise monitoring. The method consists of two modes of acoustical scene analysis, which we call static and dynamic modes, respectively: The static mode analysis examines the cumulative sound exposure distribution on the surface of the unit sphere over a long time period such as a daytime or nighttime, whereas the dynamic mode analysis estimates the trajectory of a moving sound source, which causes a sound event at the monitoring site such as an aircraft fly-over or a car passing-by, and depict it on the map. This analysis is expected to make it easier to understand the dynamic situation of the soundscape at the noise monitoring site.

1 kcie@rion.co.jp 2 i-yamada@rion.co.jp

In this paper, we make a brief report on the current status of our efforts to improve the functionality of these two modes of acoustic scene analysis. In the static mode, we are modifying the method to be repeatedly analyzed at a short-time interval, e.g., one hour, and show the result as a movie picture to facilitate understanding of the time-varying sound environment surrounding the monitoring site. On the other, in the dynamic mode, we intend to expand the function of the method to estimate the trajectory of an arbitrary sound source moving in the three-dimensional space by means of triangulation: For that purpose, we set up an auxiliary noise monitor in a proximity of the main site of unattended noise monitoring. In the previous conditions for the dynamic mode, the estimation was based on the 3D sound arrival direction and sound pressure level obtained only at the main noise monitoring site, and thus it was inevitable to assume that the sound source was a car running on the flat ground surface or an aircraft approaching along a straight path.

2. EXPANSION OF THE STATIC MODE ANALIYSIS OF ACOUSTIC SCENES

2.1. The idea behind this expansion When installing unattended aircraft noise monitors around the airport, we have to consider various factors affecting the monitoring in the selection of monitor site. One of those is the sound environment surrounding the equipment. As aircraft become quieter and quieter, the impact of environmental noise is becoming relatively larger. It must be ensured that various environmental sounds other than aircraft noise do not interfere with the measurement of aircraft noise. Even after the equipment has been installed, the influence of environmental sounds on the monitoring must always be considered. Usually, as a means of detecting single event sounds such as aircraft noise, we calculate 90 or 95 % percentile level of sound pressure levels over a time interval of certain length and use it as the threshold for detecting sound events. However, it is not sufficient for understanding characteristics of sound environment precisely at the monitoring site, because various sounds consisting of environmental noise happen to occur irregularly, and both the instantaneous sound pressure level and three-dimensional sound arrival direction of each sound varies greatly.

Therefore, it is desirable for noise monitors to be able to measure and analyze the situation of environmental noise in detail simultaneously with the measurement of aircraft noise events. Based on this idea, we proposed a means of the static mode acoustic scene analysis. In our previous presentation1), we drew a map of the sound exposure level distribution by the 3D sound arrival direction of environmental noise over a long-time interval such as daytime (16 hours) and nighttime (8 hours) and depicted a bird's-eye view of the static soundscape. However, as mentioned above, the situation of environmental noise fluctuates over time, and therefore, it is not always possible to properly evaluate fluctuations of environmental noise if the analysis time interval is fixed as daytime or nighttime. Therefore, we considered making the time interval as variable, e.g., to every 1 hour, 2 hours, ..., 24 hours, and so on. The static-mode acoustic scene analysis was modified to be repeated hourly, and to display 24 acoustic scenes per day in turn, so that we can view the change in the environmental sound source distribution at the monitoring site as if we look at a movie. We can evaluate the influence of environmental noise on the observation of aircraft noise dynamically.

2.2. Illustration of the modified static-mode acoustical scene analysis

Figure 1 shows an illustration of a result of the revised static-mode acoustic scene analysis. The figure shows a series of a bird's-eye view maps of the sound exposure level distribution of environmental noise by the 3D sound arrival direction over 12 one-hour periods. From the method of graphing in two major categories of daytime and nighttime, we have improved it so that you can dynamically see the changes during the daytime. When there are many noise events, it is not realistic to grasp the daily sound environment while checking the level variation of each noise event and the direction of sound arrival, but we thought that graphing these levels would be an easy tool to recognize the sound landscape.

Figure 1 Example of hourly static analysis results (from 0:00 to 12:00) the sound exposure level distribution of environmental noise by the 3D sound arrival direction

3. EXPANSION OF THE DYNAMIC MODE ACOUSTIC SCENE ANALIYSIS

3.1. Field experiment to examine the validity of the revised method

In our previous research, the sound source position was estimated, but for that purpose, assumptions were required from the 3D position information of the measuring instrument, actual measurement information of the arrival direction of the sound, and information such as topography and structures around the measuring instrument. For example, if the sound is estimated to be from an aircraft flying overhead, the altitude of the aircraft is assumed because the approximate distance to the touchdown point on the runway is known, or if the sound source is estimated to be on the ground, the relative height of the ground is assumed as input information for the source location estimation method, which we call dynamic analysis.

However, such preconditions cannot be established for arbitrary sound sources. Even if the positional relationship between the runway and the measurement point can be identified, it is not possible to clearly indicate the preconditions for takeoff aircraft because the flight altitude varies depending on the type and weight of the aircraft. Therefore, we decided to modify the measurement method by establishing an auxiliary measurement point in the vicinity of the measurement point, performing simultaneous measurements, and using the results to identify the position of the sound source from time to time based on the principle of triangulation. This is hereafter referred to as the revised dynamic-mode acoustical scene analysis method.

The sound source position cannot be estimated only by observation data of the direction of arrival of sound at a single point (a set of devices). To estimate the sound source location, there are two possible methods: one is to detect the distance from the sound source location to the measurement device in addition to the sound arrival direction information, and the other is to obtain the sound arrival directions at multiple points and detect the intersection of those sound arrival directions as the sound source location. We first considered the direction of sound source location estimation from the sound arrival direction information at two measurement points.

Figure 2 shows the "3D model of the dynamic analysis method without assumptions" for estimating the source position, where P A and P B are the positions of the instruments and P F is the source position. If the latitude and longitude of the PA or PB monitoring device and the orientation of the sensor are known, and the relative positions of the PA and PB (distance, altitude difference, and positive azimuth) are known, the distance and azimuth from the PA or PB can be obtained.

Figure 2 3D model for dynamic mode analysis without assumption

3.2. Current analysis of noise observations collected in the field experiment

The following is a description of an experimental system for estimating the location of a sound source using the above mentioned method. Figure 4 shows the measurement points and aircraft flight conditions. As an example of the general situation at the monitoring site, a small jet plane flies overhead at approximately 2,500 ft. The flight path on the west side of the monitoring point is a straight approach to the runway, and the horizontal distance from the monitoring point is about 1 km. The flight path to the east of the monitoring site is the turning part of the path where the aircraft turns and enters the landing path.

Figrure.4 shows the appearance of these two monitoring arrangements. In this experiment, two monitoring devices were placed at one monitoring site and are referred to as Sensor A and Sensor B. Sensor A and Sensor B are generally located in a north-south. Sensor A is a set of equipment located about 3 m above the floor on the rooftop of building, and Sensor B is another set of equipment located above the water tank on the roof of the building. This arrangement is also used to take into account the fact that there are various facilities on the rooftop, and it is not always possible to install two sets of equipment at the same height.

Since the concept of aircraft noise measurement is to "obtain measurements that are representative of the area", the monitoring points are often located on the roof of the building, avoiding shields and reflective objects. This arrangement is also based on the fact that there are various facilities on the rooftop and it is not possible to install two sets of equipment at the same height. Although the two sets of monitoring devices are independent of each other, each is clock calibrated by GPS to ensure that they do not differ by more than 0.1 second in the time direction.

The horizontal distance between the two sets of sensors should be 12 meters, and they should observe the direction of arrival of sound from the same sound source. The location of the sound source should be able to be estimated from the intersection of the directions of arrival of the sound obtained from these two sets of devices. Assuming a situation where the two sets of devices are actually installed in the same location, a confirmation experiment of this method is being conducted at a distance of approximately 12 meters.

17km to runway

Monitoring Site

Almost aircraft flights over monitoring site with about 2500ft altitude.

Figure 3 Relationship between monitoring site and flight path

Sensor B

Sensor A

Figure 4 Location outlook of microphones and sensors of our system for field experimental

3.2. Current analysis of noise observations collected in the field experiment

The following figure shows the observation situation with this experimental system. Figure 5 shows an example of data for a single noise event section where an aircraft flying northward on the west side of the measurement point was observed, and Figure 6 shows an example of data from an aircraft that turned from the east side and flew northward.

The solid line in the level variation graph on the left side of Figure 5 shows that LASmax varies by 68 dB from LAS. The graph on the right side shows the direction of arrival of sound, with the red dots representing sound arriving from above the microphone (positive elevation angle) and the gray dots representing sound arriving from below the microphone (negative elevation angle). The red dots indicate direct sound from the aircraft, while the gray dots indicate ground reflection or sound from air conditioning equipment on the rooftop.

se 36 ol A : i

The small number of gray dots (negative elevation angle) in the right-side graph in Figure 6 indicates that the system detects more direct sound arrival direction from the aircraft and a smaller percentage of sound from ground equipment, ground reflections, and other sources.

We observe such data with two sets of observation equipment. Since the microphone positions are different relative to the aircraft, which is the source of the sound, there are small differences in the direction of sound arrival, and the aforementioned model is used to process the estimation of the source position. The process of removing the effects of noise and reflected sound is currently underway.

Figure 5 Example of observation data from an aircraft flying west of the microphone

Figure 6 Example of observation data from an aircraft flying east of the microphone 4. CONCLUSIONS

This paper made a brief report on the current status of our continuing efforts to improve the functionality of the two modes, i.e., static mode and dynamic modes, of the acoustic scene analysis, which was proposed in our previous presentation at the congress of INTER-NOISE 2021.

In the static mode, we are modifying the method to be repeatedly analyzed at a short-time interval, e.g., one hour, and show the result as a movie to facilitate understanding of the time-varying sound environment surrounding the monitoring site.

On the other, in the dynamic mode, we intend to expand the function of the method to estimate the trajectory of an arbitrary sound source moving in the three-dimensional space by means of triangulation. For that purpose, we set up an auxiliary noise monitor in a proximity of the main site of unattended noise monitoring. In the previous version of the dynamic mode, the estimation was based on the three-dimensional sound arrival direction and sound pressure level obtained only at the main noise monitoring site, and thus it was inevitable to assume that the sound source was a car running on the flat ground surface or an aircraft approaching along a straight path.

We are currently analyzing the data acquired with this device and would like to report on it continuously. In the dynamic mode acoustic scene analysis to estimate the location of the sound source, we originally wanted to make it as simple as possible using auxiliary sound level meters to reduce the impact both structurally and financially, but in our current efforts, we are using two devices that can acquire the direction of sound arrival respectively. In the future, we would also like to attempt to simplify the system after achieving functional goals.

5. REFERENCES

1. Keishi Sakoda, Ichiro Yamada, and Kenji Shinohara, Sound arrival direction and acoustic scene

analysis for the monitoring of airport noise, Proceedings of INTER-NOISE 2021 .

504 004 O24