Pollution estimation system
A pollution estimation system ( 100 ) for estimating pollution level caused by exhaust gas of motor vehicles is provided. The system comprises an acoustic sensor interface ( 110 ) arranged to obtain use-data from an acoustic sensor comprising an audio sample of a motor vehicle sound and a trained exhaust gas model unit ( 120 ) arranged to receive as input the audio sample of the use-data and to apply a trained exhaust gas model to the received audio sample to produce an estimated pollution level associated with the received audio sample. The trained exhaust gas model has been obtained by training an exhaust gas model on multiple training items using a machine learning algorithm, the multiple training items comprising multiple audio samples of motor vehicle sounds obtained from one or more acoustic sensors and associated pollution levels.
1. A pollution estimation system for estimating pollution level caused by exhaust gas of motor vehicles, the system comprising:
an acoustic sensor interface arranged to obtain use-data from connected lighting devices having an acoustic sensor comprising an audio sample of a motor vehicle sound in an area, the connected lighting devices and/or the acoustic sensor comprising identifying information;
a processor comprising a trained exhaust gas model circuit arranged to receive as input the audio sample of the use-data and to apply a trained exhaust gas model to the received audio sample to produce an estimated pollution level caused by exhaust gas of motor vehicles associated with the received audio sample, wherein the trained exhaust gas model comprises a plurality of parameters and an algorithm configured to apply the plurality of parameters to the audio sample input;
wherein the trained exhaust gas model has been obtained by training an exhaust gas model on multiple training items using a machine learning algorithm, the multiple training items comprising multiple audio samples of motor vehicle sounds obtained from one or more acoustic sensors and multiple measured pollution levels of motor vehicles obtained from one or more pollution sensors, wherein each measured pollution level is associated with each audio sample of the multiple audio samples such that each measured pollution level is paired with each audio sample of the respective multiple audio samples based on the acquired data; and
wherein the processor is configured to output the estimated pollution level and/or the identifying information when the estimated pollution level is above a pollution threshold.
2. The pollution estimation system as in claim 1 , wherein:
the exhaust gas model comprises a classification model and at least one regression model, a list of tags being defined for the exhaust gas model, each one of the regression models being associated with a tag of the list of tags, the classification model being arranged to predict a tag from the audio sample of the motor vehicle sound, and
the trained exhaust gas model circuit is arranged to:
apply the trained classification model to the audio sample of the use-data to obtain a predicted tag from the list of tags,
apply a trained regression model of the at least one regression model associated with the predicted tag to the audio sample of the use-data to produce the estimated pollution level caused by exhaust gas,
the multiple training items comprising multiple audio samples of motor vehicle sounds, associated pollution levels, one or more associated tags, the tags being selected from the list of tags,
the exhaust gas model having been obtained by:
training the classification model on the multiple audio samples of motor vehicle sounds and the associated tags of the multiple training items, thus obtaining a trained classification model arranged to predict a tag from an audio sample of a motor vehicle sound, and
training each particular regression model of the at least one on the audio samples of the multiple audio samples associated with the tag associated with the particular regression model, thus obtaining at least one arranged to estimate a pollution level from an audio sample of a motor vehicle sound of a particular tag.
3. The pollution estimation system as in claim 2 , wherein the exhaust gas model comprises multiple regression models, at least two tags of the list of tags being associated with different regression models of the multiple regression models.
4. The pollution estimation system as in claim 2 , wherein:
at least two tags of the list of tags indicate a fuel type of motor vehicle, said two tags distinguishing at least between motor vehicles using gasoline and diesel fuel, and/or
at least two tags of the list of tags indicate a current use of a motor vehicle, said two tags distinguishing at least between motor vehicles running stationary and running non-stationary.
5. The pollution estimation system as in claim 1 , wherein the exhaust gas model circuit is arranged to produce multiple estimated pollution levels of at least two different pollutants.
6. The pollution estimation system as in claim 1 , further comprising:
a sensor circuit comprising:
the acoustic sensor arranged to obtain the audio sample of the motor vehicle sound as the use-data, and
a camera arranged to capture an image of the motor vehicle if the estimated associated pollution level is above the pollution threshold.
7. The pollution estimation system as in claim 1 , comprising an outdoor luminaire, wherein the outdoor luminaire comprises the acoustic sensor arranged to obtain the audio sample of the motor vehicle sound as the use-data.
8. The pollution estimation system as in claim 1 , comprising a compression circuit arranged to reduce the size of the audio sample of the motor vehicle sound before transmitting the audio sample from the acoustic sensor to the trained exhaust gas model unit.
9. The pollution estimation system as in claim 1 , wherein the multiple training items are obtained by:
continuously measuring audio during a training interval from the acoustic sensor at a roadside location;
continuously measuring pollution levels during the training interval from the one or more pollution sensors located at the motor vehicle proximate the roadside location; and
obtaining multiple audio samples of motor vehicle sounds and associated pollution levels as parts of the continuously measured audio and corresponding parts of the continuously measured pollution levels.
10. The pollution estimation system as in claim 1 , comprising:
a machine learning circuit arranged to obtain from one or more acoustic sensors multiple motor sounds of multiple motor vehicles in use, and associated pollution levels of the multiple motor vehicles in use, thus obtaining multiple training items comprising multiple audio samples of motor vehicle sounds obtained from one or more acoustic sensors and associated pollution levels, and
train an exhaust gas model on the multiple training items using a machine learning algorithm, thus obtaining a trained exhaust gas model.
11. The pollution estimation system as in claim 1 , wherein the exhaust gas model circuit is arranged to compute one or more features from the received audio sample, and to apply the trained exhaust gas model to the computed features.
12. A pollution estimation method for estimating a pollution level caused by exhaust gas of motor vehicles, the method comprising:
obtaining use-data from connected lighting devices having an acoustic sensor comprising an audio sample of a motor vehicle sound in an area, wherein the acoustic sensor and/or the connected lighting devices comprise identifying information;
applying a trained exhaust gas model to the received audio sample to produce an estimated associated pollution level caused by exhaust gas of motor vehicles, wherein the trained exhaust gas model comprises a plurality of parameters and an algorithm configured to apply the plurality of parameters to the audio sample input;
wherein the trained exhaust gas model has been obtained by training an exhaust gas model on multiple training items using a machine learning algorithm, the multiple training items comprising multiple audio samples of motor vehicle sounds obtained from one or more acoustic sensors and multiple measured pollution levels of motor vehicles obtained from one or more pollution sensors, wherein each measured pollution level is associated with each audio sample of the multiple audio samples such that each measured pollution level is paired with each audio sample of the respective multiple audio samples based on the acquired data; and
outputting the estimated associated pollution level and/or the identifying information when the estimated associated pollution level is above a pollution threshold.
13. The pollution estimation method as in claim 12 , the method comprising:
obtaining, from one or more acoustic sensors, multiple motor sounds of multiple motor vehicles in use, and associated pollution levels of the multiple motor vehicles in use, thus obtaining the multiple training items comprising multiple audio samples of motor vehicle sounds and associated pollution levels, and
training the exhaust gas model on the multiple training items using the machine learning algorithm, thus obtaining the trained exhaust gas model.
14. Computer program for estimating a pollution level from exhaust gas of motor vehicles arranged to cause a processor to perform the method as claimed in claim 12 .
15. A computer program as in claim 14 embodied on a computer readable medium.
16. A roadside pollution estimation device configured to generate a pollution level estimate caused by exhaust gas of a motor vehicle, the device comprising:
an acoustic sensor configured to record an audio sample of a motor vehicle sound;
a processor configured to receive the obtained audio sample, and further configured to apply a trained exhaust gas model to the received audio sample to generate a pollution level estimate caused by exhaust gas for the obtained audio sample, wherein the exhaust gas model comprises a classification model and a regression model associated with each tag of a list of tags defined for the exhaust gas model, the classification model being arranged to predict a tag of the list of tags from the audio sample of the motor vehicle sound, and each regression model being arranged to generate the pollution level estimate based on the tag predicted by the classification model; and
a communication element configured to communicate the generated pollution level estimate;
wherein the trained exhaust gas model is generated via a machine learning algorithm trained on multiple training items, the multiple training items comprising multiple audio samples of motor vehicle sounds obtained from one or more acoustic sensors and multiple measured pollution levels of motor vehicles obtained from one or more pollution sensors, wherein each measured pollution level is associated with each audio sample of the multiple audio samples such that each measured pollution level is paired with each audio sample of the respective multiple audio samples based on the acquired data.
17. The roadside pollution estimation device as in claim 16 , wherein:
at least two tags of the list of tags indicate a current use of a motor vehicle, said two tags distinguishing at least between motor vehicles increasing speed and decreasing speed.