1. ANOMALOUS NOISE EVENTS CONSIDERATIONS FOR THE COMPUTATION OF ROAD TRAFFIC NOISE LEVELS : THE DYNAMAP'S MILAN CASE STUDY
F. Orga (1), R. M. Alsina-Pagès (1), F. Alías (1), J. C. Socoró (1), G. Zambon(2), R. Benocci(2)
1) GTM - Grup de recerca en Tecnologies Mèdia, La Salle - Universitat Ramon Llull, Barcelona
2) Department of Earth and Environmental Sciences, University of Milano Bicocca, Milan

2. Outline Introduction What is an ANE? And ANED?
Milan Scenario description
Analysis of the ANE
Conclusions

3. 1-Introduction (1)
LIFE-DYNAMAP (Dynamic Acoustic Mapping - Development of low cost sensors networks for real time noise mapping) – ( ). Project goals:
Implement Environmental Noise Directive 2002/49/EC
Acoustic maps based on low-cost smart grids to ease and perform its updating.
Two pilots areas: Rome (highway portals) and Milan (district 9)
Development of an Anomalous Noise Event Detection (ANED) algorithm, that gives support to automatic data record and processing for the acoustic maps, helping in detect noise that only is caused by road traffic.

4. 1-Introduction (2)
The goal of the work we present today is:
Identify the impact of ANE in several examples of real-life measurements.
Identify if the several measurement places in Milan present homogeneous results in terms of ANE typology or they should be clustered.
* The measured roads could be classified into 2 clusters for their performance and ANE typology, following the clusters that Bicocca had already defined [1]
[1] G. Zambon, R. Benocci, G. Brambilla. Cluster categorization of urban roads to optimize their noise monitoring. Environmental Monitoring Assessment,(2016) 1, 88: 26.

5. 2. What is an ANE? (1) ANE stands for Anomalous Noise Event
END 2002/49/EC states that the noise impact of each main source should be determined and reported separately -> other noise sources should not be considered
Our goal is to evaluate the road traffic noise
The automatic ANE identification and labelling became one of the goals of the project, in order to minimize the ANE impact to the equivalent measured level

6. 2. What is an ANE? (2)

7. 2. What is an ANE? (3)
ANE stands for Anomalous Noise Event

8. 2. What is ANED? (4)
Supervised baseline technique

9. 2. ANED Requirements (5) Dataset to train the system
Road Traffic Noise (RTN) raw data
Anomalous Noise Events (ANE) raw data
Less than 3% of data in Rome
Around 12% of data in Milano
Quality, diversity, amount and distribution of existing ANE determines the ANED performance

10. 3. Milan Scenario Description
2233 road stretches
belonging to all different types of urban roads:
principal arterial roads
collector arterial roads
local roads

11. 3. Milan Scenario Description
By means of a cluster analysis, based on the Leq roads trend, we obtain two clusters clearly distinct
Box plots of the traffic flow rate at rush hour for the two cluster.

12. 3. Milan Scenario Description
The monitoring devices have to be provided of ANED
In order to check the feasibility of deleting ANE, preliminary tests sites with different acoustic characteristic were selected:
Sites embracing only primary road
Sites with additional crossings
Sites with railway lines nearby
Complex mixed scenarios

13. 4. Analysis of the ANE (1)
19 different types of ANE in DYNAMAP

14. 4. Analysis of the ANE (2)
ANE occurences for each cluster

15. 4. Analysis of the ANE (3)
ANE duration boxplots

16. 4. ANE impact on Leq 5min (1)
Evaluation of the Laeq 5 mins by means of interpolation of the ANE detected:
It is necessary to describe in terms of Leq 1sec which is the calculus associated to the Leq5mins. When an ANE is found, an interpolation is conducted with the value of Leq1sec. When the ANE has low SNR, the interpolation value is higher. When the SNR has high value, the interpolation is lower (which is the correspondance to reality).

17. 4. ANE impact on Leq 5min (2)
We measure the impact at 5min because of the project specifications
Two 20’ projects:
Left: impact of non-salient ANE (<1 dB)
Right: impact of a salient ANE (>6 dB)
This two comparisons are really clear of what is happening.
The first project has not many ANE with high saliency (high SNR), so the difference every 5 mins is low, máximum 1 dB
The second project has at least one point (10’) where there is a long and salient ane (a siren) that makes the difference between the evaluation with or without ANE is really different (around 6db)
For Giovanni: i put this two figures because tehy are very clear, and their results present the same values tan the projects we are working now. I have just put one new figure, in next slide.

18. 4. ANE impact on Leq 5min (3)
Figure is a project of 15 minutes, evaluated every 5 mins
Blue curve represents the Leq5mins including all ANE
Red curve represents the Leq5min without ANE
All other curves are analysis over the SNR of the ANE
Up to 2.5dB difference in 5 mins (min 10)
This is a new figure from a project in Milano. We have several curves here. Most of them correspond to the study of threshold of SNR in the ANE, to evaluate the difference in the impact of the ANE and Leq5min.
The value of Leq5min with all the ANE (the máximum value of the Leq) is nearly 3db over the value of the ground truth (that we can consider the previous one or just the one showing 0dB as we talked last week). You can use this slide to compare the concept of ANE (which is independent of its SNR) to show that the importance of the SNR is crucial to know which of the ANE have to be supressed (interpolated) and which have to be integrated as background noise.

19. 5. Conclusions
Excluding the ANE has a positive impact, as the Leq5min obtained is lower and more realistic evaluation of the noise traffic in the street
The two clusters found by Bicocca team are correlated with the occurrences of ANE in the recorded measurements
Future work and analysis
ANE duration and SNR study and their impact in Leq
Critical ANE definition in terms of impact at Leq 5mins
Final ANED adjustments will be conducted depending on the previous results of the analysis