This application claims the benefit of European Patent Application No. 16159645.7, filed on Mar. 10, 2016, which is hereby incorporated by reference herein.
The invention relates to a pollution estimation system, a pollution estimation method, a computer program, and a computer readable medium.
Motor vehicle emissions contribute to air pollution and are a major ingredient in the creation of smog in cities. It is estimated that 15,000 hospital admissions are caused annually by diesel exhaust pollution in the United States (See “An Analysis of Diesel Air Pollution and Public Health in America. (v. 1.3)”, Revised June 2005, by the CATF, page 8). There is therefore a desire to monitor pollution caused by vehicle exhausts, especially in cities. Through the monitoring insight may be obtained where pollution through vehicle exhaust is a particular problem. This information may then be used, say, for planning purposes, e.g., when planning new infrastructure in the city, or to regulate traffic, etc.
Pollution to which vehicle exhaust gas contributes includes the following: carcinogens, mono-nitrogen oxides (NOx) including nitric oxide (NO) and nitrogen dioxide (NO2), hydrocarbons (HC), carbon monoxide (CO), carbon dioxide (CO2), fine particles in particular diesel particulate.
Monitoring of vehicle exhaust gas is challenging as it requires sensing close to the source. Ideally, the sensing would be near the exhaust pipe of vehicles itself. Although this is possible to do for test cars, it is not suitable for large scale monitoring. Moreover, the potentially large number of different pollutant makes sensors also relatively expensive, if more than one type of pollution needs monitoring.
It would be advantageous to have an improved method of monitoring pollution caused by motor vehicle exhaust.
A pollution estimation system for estimating pollution level caused by exhaust gas of motor vehicles is provided. The system comprises
By taking an audio fragment of the sound of a motor vehicle, say a car, the trained exhaust gas model unit can estimate the pollution caused by that motor vehicle. In this way, the pollution estimation system does not require a pollution sensor to estimate pollution levels. In fact acoustic sensors may be used for multiple purposes and also function to estimate pollution levels.
The exhaust gas model may be, e.g., a neural network. The estimation system may be improved by a hybrid model comprising both a classifying model as well as a regression model. For example, in an embodiment, the exhaust gas model comprises a classification model and at least one regression model. A list of tags is 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 is arranged to predict a tag from an audio sample of a motor vehicle sound, and
The exhaust gas model may have been obtained by
In an embodiment, 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. It was found that a classifier can distinguish between the sounds of a car on diesel and one running on gasoline. A regression model specialized for diesel or gasoline then translates the sound to an estimate.
In an embodiment, the system comprises an outdoor luminaire, e.g., an outdoor lighting pole, wherein the outdoor luminaire comprises an acoustic sensor arranged to obtain an audio sample of a motor vehicle sound as use-data. For example, the system may comprise multiple acoustic sensors distributed in an area, e.g., in lighting poles.
In an embodiment, the pollution estimation system is connected to a traffic management system, the pollution estimation system sending a warning message, if pollution levels exceed a threshold, to the traffic management system causing the traffic management system to adjust traffic flow.
The pollution estimation system is an electronic system. The exhaust model unit is an electronic device and may be implemented on a computer.
The pollution estimation method described herein may be applied in a wide range of practical applications. Such practical applications include localizing regions where pollution levels are high, controlling traffic, and the like.
A method according to the invention may be implemented on a computer as a computer implemented method, or in dedicated hardware, or in a combination of both. Executable code for a method according to the invention may be stored on a computer program product. Examples of computer program products include memory devices, optical storage devices, integrated circuits, servers, online software, etc. Preferably, the computer program product comprises non-transitory program code stored on a computer readable medium for performing a method according to the invention when said program product is executed on a computer.
In a preferred embodiment, the computer program comprises computer program code adapted to perform all the steps of a method according to the invention when the computer program is run on a computer. Preferably, the computer program is embodied on a computer readable medium.
Further details, aspects and embodiments of the invention will be described, by way of example only, with reference to the drawings. Elements in the figures are illustrated for simplicity and clarity and have not necessarily been drawn to scale. In the Figures, elements which correspond to elements already described may have the same reference numerals. In the drawings,
While this invention is susceptible of embodiment in many different forms, there are shown in the drawings and will herein be described in detail one or more specific embodiments, with the understanding that the present disclosure is to be considered as exemplary of the principles of the invention and not intended to limit the invention to the specific embodiments shown and described.
In the following, for the sake of understanding, elements of embodiments are described in operation. However, it will be apparent that the respective elements are arranged to perform the functions being described as performed by them.
Further, the invention is not limited to the embodiments, and the invention lies in each and every novel feature or combination of features described above or recited in mutually different dependent claims.
There is a need to obtain insight in the pollution in an area, say in a city, which does not require equipping each car with a pollution sensor. In fact, it is also preferred if pollution sensors where not or only limitedly needed at all. Embodiments described below use acoustic sensors, say mounted in light poles, to record vehicle sounds. From the sounds of the vehicles the pollution is then estimated. This provides a low cost way to provide quick insight in the pollution in an area. Especially relative pollution, e.g., how much worse pollution is in one area compared to another area can be quickly obtained in this manner. Compared with distributing a large number of pollution sensors the costs are also relatively low. Furthermore, in an embodiment, quite different types of pollutions, say mono-nitrogen oxides and diesel particulate, may be estimated from a single sensor, in this case an acoustic sensor. In the context of this document, pollution refers to pollution caused by exhausts fumes of motor vehicles.
Interestingly, motors producing high exhaust gas also produce loud noise at that moment. An embodiment is proposed in which a connected lighting infrastructure is equipped with acoustic sensors that capture sound of motors exhausting gas. As the location of each light point is known, the location and tracking of vehicles producing high volume of exhaust gas can be done. Acoustic sensor may also be installed in different locations than light poles.
Shown in graph 600 is a horizontal axis 620. Horizontal axis 620 shows the vehicle speed of an auto. The upper axis shows the speed in kilometers per hour, the bottom axis shows the speed in miles per hour. A vertical left axis 610 shows the A-weighted sound-level in dBA.
The sound-level is obtained as a single vehicle passes by a microphone 15 meters (50 feet) to the side, its sound level rises, reaches a maximum, and then falls as the vehicle recedes down the roadway. The maximum A-weighted sound level during the pass by is called that vehicle's noise-emission level. We refer to the “Federal Highway Administration's Traffic Noise Model (FHWA TNM®), Version 1.0—Technical Manual” of the US Department of Transportation for the exact computation of A-weighted noise levels. Solid line 640 shows a relationship between the speed and noise-emission level of an auto vehicle. Note that the noise level increases with the speed.
Pollution estimation system 100 comprises an acoustic sensor interface 110 arranged to obtain use-data from an acoustic sensor 210. The use data comprises an audio sample of a motor vehicle sound. An audio sample may also be referred to as an audio fragment. For example, pollution estimation system 100 may be implemented in a computer, e.g., a server. The audio interface 110 may be a computer network connection, e.g., an internet connection. For example, a connection may be established between acoustic sensor 210 over a 3G connection and/or a wired a connection, etc. Acoustic sensor 210 may be a microphone. Acoustic sensor 210 and pollution estimation system 100 may be two separate systems located at different places. However, different distributions are also possible. For example, pollution estimation system 100 and one or more acoustic sensors may together form a system.
Pollution estimation system 100 may be integrated in a single device, e.g., a computer. Pollution estimation system 100 may also be implemented in a distributed fashion. For example, pre-processing, e.g., feature extraction (see below) may be performed at acoustic sensor 210 while the rest of the pollution estimation takes place at a further device implementing the rest of pollution estimation system 100. For example, a pollution estimation device may comprise acoustic interface 110 and trained model unit 120.
For example, interface 110 may be internal to an estimation device receiving audio samples from an acoustic sensor. For example, interface 110 may be internal to a sensor unit receiving audio samples from an acoustic sensor. In the latter case the sensor unit may perform part or all of the processing of the exhaust gas model.
Acoustic sensor 210 may be arranged to record sound continuously, e.g., for longer than the duration of an audio sample. From the longer recording, the audio sample may be selected. Alternatively, acoustic sensor 210 may be arranged to record an audio sample occasionally; e.g., at regular intervals; e.g. at the request of pollution estimation system 100, etc. An audio sample, may be a few seconds long, say 5 seconds, or 10 seconds, etc. A longer audio sample may be an hour long, a day, or non-stop, etc.
Pollution estimation system 100 comprises a trained exhaust gas model unit 120 arranged to receive as input the audio sample of the use-data and to apply trained exhaust gas model 120 to the received audio sample to produce an estimated pollution level associated with the received audio sample.
For example, the exhaust gas model may comprise a large number of parameters, e.g., stored in an electronic storage, say in a memory, and an algorithm that applies said parameters to the inputs. The algorithm may be predetermined, say determined by the type of machine learning and system designer. The parameters may be selected by a machine learning algorithm.
Trained exhaust gas model 120 has been previously obtained by training an exhaust gas model on multiple training items using a machine learning algorithm. In this way, an estimation of pollution, or of a pollution increase, may be obtained from audio data, i.e., the audio data obtained from acoustic sensor 210, even if no pollution sensor is available.
It is noted that the audio sensors may also be used for other purposes. For example, acoustic sensor 210 and/or further acoustic sensors used by system 100 may be used for gunshot detection. A distributed array of acoustic sensors may detect the gunshot blast at slightly different times, which may be used to calculate the location of the origin of the discharge. In an embodiment, the acoustic sensors are arranged with a clock to specify when the audio was recorded.
Pollution estimation system 100 may optionally comprise a machine learning unit 130. Machine learning unit 130 has access to a training database 132. Training database 132 may be located in pollution estimation system 100, e.g., at the same location as exhaust gas model unit 120. For example, training database 132 may be stored in a server, in the cloud etc.
Training database 132 stores multiple training items. The multiple training items comprise multiple audio samples of motor vehicle sounds obtained from one or more acoustic sensors and associated pollution levels.
To construct the multiple training items multiple motor sounds of multiple motor vehicles in use have been obtained from one or more acoustic sensors, e.g., sensors like acoustic sensor 210. The multiple motor sounds, e.g., audio samples, are associated with pollution levels of the multiple motor vehicles in use. In addition to an audio sample, and a pollution level, a training item may also comprise one or more tags. The tags give information on the type of motor, type of use etc. Tags provide additional training data to the model. Other relevant information may be recorded in database 132 as well, e.g., time of day, date, measurement location, etc. Some of the other relevant information may be used as additional inputs; for example, information on environmental conditions, such as, e.g., temperature, wind speed, and humidity.
The tags in the training items may be selected from a list of tags. In an embodiment, at least two tags of the list of tags indicate a fuel type of the motor vehicle, said two tags distinguishing at least between motor vehicles using gasoline and diesel fuel. A further tag may be electric or LPG, e.g. gas. For example, the list of tags indicating a fuel type of the motor vehicle may be {diesel, gasoline}, {diesel, gasoline, electric}, or, {diesel, gasoline, electric, gas}, etc.
In an embodiment, 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. A further use may be increasing speed, decreasing speed, running at constant speed, etc. For example, the list of tags indicating a use type may be {running-stationary, running-non stationary}, or, {running-stationary, constant speed, increasing speed, decreasing speed}, etc.
The use of tags is optional. In an embodiment, only tags of certain types may be used, e.g., only tags that indicate fuel type or motor type. The tags may be added to training items on an availability basis. On the other hand, in an embodiment, each training item has assigned to it all appropriate tags. A list of tags may have the tag ‘unknown’.
For example, in an embodiment, database 132 may comprise a table, containing pairs of an audio-sample reference, e.g., a pointer, a filename etc, and a pollution level, and one or more tags. For example, a table may look like this:
The “Audioref_ptr” refer to an audio sample. The values refer to a pollution level in a suitable unit and using a suitable measurement method. The tags may refer to information to the type of motor or use, etc. In this case tags are assigned from the two lists {gasoline, diesel, electric} and {non-stationary, stationary}. Two lists of tags may be combined in a single tag list. For example, a tag may be used that indicates both fuel type and use type. For example, the list of tags may be {non-stationary+gasoline, non-stationary+diesel, non-stationary+electric, stationary+gasoline, stationary+diesel, stationary+electric}. The above values are examples, for specific cars, and a specific pollutant. In an embodiment, values are provided for different or multiple pollutants.
Machine learning unit 130 is arranged to train an exhaust gas model on the multiple training items using a machine learning algorithm, thus obtaining a trained exhaust gas model.
There are several machine learning algorithms that may be used including classifying and regression algorithms. Suitable classifiers include neural networks, support vector machines, decision tree and random forests. Regression may be regression tree, polynomial regression. Conventional machine learning algorithms may be used for machine learning unit 130. Below advantageous hybrid architectures will be discussed that is also capable of using tags. Such tags may significantly reduce the required number of training items and/or increase the accuracy of the system for a given amount of training data.
Once exhaust gas model unit 120 is trained, machine learning unit 130 is not needed anymore. So database 132 and machine learning unit 130 are optional in pollution estimation system 100. On the other hand, to train the exhaust gas model no acoustic sensor interface 110 is needed, if a database 132 is available. An embodiment of a machine learning system comprises exhaust gas model 120, database 132 and machine training unit 130.
Training data may be obtained in several ways. For example, audio samples of common car types may be obtained, e.g., measured using an acoustic sensor. For example one may put a motor vehicle, say a car, on a roller bank and obtain the audio sample. The audio sample may be a few seconds, e.g., 5 seconds. An audio sample may be longer than a training sample. Together with obtaining the audio sample, also exhaust pollution information may be obtained, e.g., by measuring it. The audio recording and pollution sensoring may be run continuous. While the car is run through a sequence of different operations, different pollution values are measured. In this way, a large audio sample, say of an hour long, and a corresponding record of the pollution is obtained. Later the large audio sample may be cut up into audio samples say of 5, 6, seconds etc. For example, a moving window of a pre-determined audio sample size may be used. The sequence may be standardized and used for multiple vehicles. The location and type of pollution sensor may also be standardized. For example, the pollution sensor may be located at the exhaust pipe of the car. This will lead to a pollution estimation system that estimates pollution at the exhaust pipe, even though the acoustic sensor is located further away.
If tags are used, the audio sample and pollution pairs may also be associated to one or more tags. For example, a training item may be associated with a type of car, e.g., bus/car, and a use type, increase gas, stationary, etc. For example, the training data may be represented as a table with in a first column audio samples, a second column a pollution level, and a third column a car classification.
Training data may be also obtained by placing the pollution sensor further away from the exhaust pipe. Training data may even be obtained by sound recording at a road used for motor vehicles. A human may spot and record tags, say auto types. A sensor may measure pollution at the roadside. As the pollution sensor is not directly located near the exhaust pipe in this embodiment, a reduced accuracy must be accepted. A time delay may be introduced between measuring the audio sample and the measuring of the pollution. For example, if the time delay is 1 minute, an audio sample measured at time t=0 will be associated with pollution levels measured at t=1 minute. In any embodiment, it is possible to measure an increase in pollution levels rather than an absolute pollution level; in this embodiment this possibility is especially useful, as there may be background pollution. Furthermore, in this embodiment, the training items may be provided with additional factors that impact distribution of the pollution level between the motor vehicle and the sensor, say humidity. The humidity may be provided to the exhaust model as an additional input, both during training and use.
In an embodiment, road side training data is used and audio samples are much longer than 5 seconds, say 1 hour or more. The 1 hour audio sample may be aggressively reduced in size and/or feature extraction may be done at the sensor. For example, the audio sample may be divided into, e.g., 20 second long periods, of which energy is computed. In combination the model may predict pollution increase of a road size sensor rather than absolute pollution values.
The model may also be trained on samples having sound of multiple vehicles in them. Such combined samples may be obtained at a roadside location. Combined samples may also be produced from a roller bank. For example, given multiple training items of a single car, the audio may be mixed to obtain sample corresponding to multiple cars. The model may instead be trained only on audio samples of single cars; during use the model may be applied to single cars passing the sensor.
In an embodiment, the multiple training items are obtained by continuously measuring audio during a training interval from an acoustic sensor at a roadside location. For example, the training interval may be longer than one audio sample; for example the training interval may be at least 500 times longer than the duration of a single audio sample; for example, the training interval may be an hour or a day using audio samples of 5 seconds. The embodiment further comprises continuously measuring pollution levels during the training interval from a pollution sensor at the roadside location. From the recorded pollution levels and the recorded sound, multiple audio samples of motor vehicle sounds may be selected together with a corresponding selection from the pollution level; the latter may include a delay to compensate for the time the pollution takes to travel from the vehicle exhaust to the sensor.
The training items may be used with a monolithic model, say one neural network or support vector machine that takes as input an audio sample, and/or features computed from the audio samples (see below) and produces as output a pollution level estimate. Training data of this kind is especially suited for use with a monolithic model, say a large neural network, say, employing deep learning. Such a system may also use tags, e.g., as an additional input to the model, but it is not necessary. An advantage of, say, a deep neural network is that relatively little investment in features and model architecture is required. Below embodiments are described with combine different types of trainable models, and which are capable of using tags advantageously.
Each tag is associated with a regression model of the multiple regression models. For example, diesel may be associated with regression model 124.1; gasoline with 124.2 and electric with 124.3. In an embodiment, a plurality of tags from the list of tags are associated with the same regression model. This option allows the model to be extended later with more detailed regressions models. For example, this is useful as more training data becomes available.
Classification model 121 is arranged to predict a tag from an audio sample of a motor vehicle sound. Exhaust gas model unit 120 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. For example, classification model 121 is provided with the audio sample of a motor vehicle sound from the use-data and predicts a tag, or one or more tags, associated with the audio sample. For example, if classification model 121 were presented with audio sample ‘Audioref_ptr_31534’ of the previous example, then classification model 121 would be expected to produce the tag gasoline.
The classification model may produce a probability measure for each of the tags in the list of tags. For example, classification model 121 may produce as output that tag gasoline has a probability of 78%, diesel of 15%, and electric of 7%.
For example, classification model 121 may be a neural network. The neural network may have an input layer comprising multiple input-nodes. At least part of the input-node may receive audio values from the audio sample possibly after preprocessing, e.g., Fourier analysis, frequency reduction, etc. We refer to “Audio chord recognition with recurrent neural networks”, by Boulanger-Lewandowski, et al. for appropriate neural network architecture. Note that the although the cited paper is completely unrelated to pollution estimation or to motor vehicles in general, the neural networks for processing audio may be used for exhaust gas model 120, and in particular for classifier 121.
Exhaust gas model unit 120 is arranged to apply the trained regression model associated with the predicted tag to the audio sample of the use-data to produce an estimated pollution level. For example, in the above example, exhaust gas model unit 120 might take gasoline as the predicted tag, e.g., because it has the highest probability and apply the corresponding regression model 124.2.
For example, exhaust gas model unit 120 may comprise a selector 122 arranged to receive the output of classification model 121, and determine a suitable regression model therefrom. For example, selector 122 may be arranged to determine a predicted tag from said output and select and apply the regression model that is associated with the predicted tag.
In an embodiment, each of the regression models is applied and the resulting predictions are averaged using as weights the probability of each tag. For example, said weighting may be performed by selector 122.
This model is a hybrid model containing both a classification model and a regression model. To train the hybrid model, multiple training items may be used comprising multiple audio samples of motor vehicle sounds, associated pollution levels, and one or more associated tags, the tags being selected from the list of tags.
The classification model is trained by training unit 130 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. Each one of the regression models is trained by training unit 130 on the audio samples of the multiple audio samples associated with the tag associated with the particular regression model, thus obtaining multiple trained regression models arranged to estimate a pollution level from an audio sample of a motor vehicle sound of a particular tag.
It is an insight of the inventors that different types of vehicles have a different type of relationship between sound and pollution. By isolating the classification part of the problem, this part can be optimized separately. For example, the classifier may be chosen to be a neural network or support vector machine, etc. The regression models may be regression tree, polynomial regression, etc.
In an embodiment, the classification model 121 is a neural network and one or more regression models are polynomials regression models.
If desired the multiple regression models may be combined in a single model. One such embodiment is shown in
In an embodiment, the exhaust gas model unit comprises a feature extractor 126. Feature extractor 126 computes values from an audio sample that are expected to be correlated with the desired outcome of the exhaust gas model unit. For example, in
For example, features may be obtained from the audio sample,
A feature may be a single number, say the total amount of energy, but may also be multiple numbers, e.g., the frequency analysis of the sample.
It was found that pollution levels and sound volume correspond reasonably well, but that exact nature of the correspondence depends on the type of motor vehicle. The embodiment shown in
In an embodiment, the exhaust model is restricted to recognizing only vehicles that are known to produce large amounts of pollution gas. For example, a tag ‘ignore’ could be introduced, such a tag could be coupled to an empty regression model that simply outputs a default value, say an average pollution level, or the corresponding audio sample could be ignored altogether.
In an embodiment, a single regression model is used. For example, the list of tags may comprise two tags: an ignore-tag and a tag that identifies a high polluting type of vehicle. For example, the list of tags may be {diesel, ignore}. The ignore tag is applied to all non-diesel vehicles, and the diesel tag is associated with sounds of diesel vehicles. Only if the classification model predicts the diesel tag, is the regression model used. The latter is trained to estimate pollution for diesel vehicles. If the ignore tag is predicted, the model may generate a corresponding signal, or may produce a default average pollution value, etc.
In an embodiment pollution estimation system 100 is distributed over multiple locations, for example feature extractor 126 is located at the acoustic sensor but classification and regression models 121 and 124 are located in a server. In one embodiment, the classification model also does not receive the full audio samples, only extracted features. The latter option can be efficiently combined with the distributed embodiment, as it saves bandwidth.
In embodiments, the acoustic sensor may be implemented in a sensor unit. Several types of sensor units are shown in
For example, processor 230 may be arranged as a feature extractor, like feature extractor 126. If only extracted features are transmitted, a significant reduction in bandwidth may be obtained.
Each one of the sensors units may be combined with a transmission unit (not separately shown in
In this document, the term lighting pole refers to the pole itself together with the light fixture at the top of the pole and additional optional driving circuitries provided in the pole and/or in the light fixture at the top of the pole. For lighting pole 330 distinct elements have been indicated, namely the light fixture 340, a pole 342 and a construction element 341 at the top of the pole which comprises, for example, a driving circuitry for driving the lamp in the light fixture 340.
In an example, the system 100 of
In an example, each one of the lighting poles 331-335 are provided with an acoustic sensor and are provided with a transmitter, say a wireless transmitter. Each sensor in the lighting poles 331-335 has an associated surface that is a portion of the street that is close to and illuminated by the respective lighting poles 331-335. For example, lighting pole 335 may detect sound produced by a car or other motor vehicle passing underneath lighting pole 335. An audio sample of the sound of the car is obtained by the acoustic sensor and transmitted wirelessly to audio interface 110. Additional information may include current time and possibly other sensor measurement, e.g., humidity sensors, wind sensors etc.
In an embodiment, acoustic sensors in lighting pole along the roads in cities recognize sound of motors that emit high volumes of exhaust gas into the air and optionally locate such vehicles. For example, an identification of the lighting pole or of the acoustic sensor that made the recording may be transmitted together with the audio sample.
In an embodiment one could use the system to estimate overall pollution levels in an area (e.g. city). In an embodiment, the pollution estimation system is connected to a traffic management system 140. This option is illustrated in
Knowing the areas with high air pollution and/or roads on which vehicles with high emission of exhaust gas are not allowed, the system can warn the authorities about location and time of such a rule violation. The authorities can also use the information to identify the source of most air pollution and use it to shape city policies. Having the pollution estimates, various uses present themselves. For example
For example to make traffic smoother one may place traffic signals on a coordinated system so that drivers encounter a progression of green lights (also known as a ‘green wave’). Cutting speed can also significantly reduce emissions of other pollutants, particularly reducing NOx and particulate matter (PM) output from diesel vehicles.
In an embodiment, the model separately warns for motor vehicles that run stationary (also known as running idle). Idle detection is important as it represents an easy way to reduce emission. For example, when location are identified where cars often idle, warning signs may be presented to reduce the amount of time drivers idle their engines. Idle detection is possible having the classification model 121 predict a stationary and non-stationary tag. Idle detection may also be done standalone without prediction pollution levels. An example of such a system is an idle detection system for detecting motor vehicles running stationary. The system comprises
Typically, the device 100 and 201-204 each comprise a microprocessor (not separately shown) which executes appropriate software stored at the devices; for example, that software may have been downloaded and/or stored in a corresponding memory, e.g., a volatile memory such as RAM or a non-volatile memory such as Flash (not separately shown). Alternatively, the devices 100 and 201-204 may, in whole or in part, be implemented in programmable logic, e.g., as field-programmable gate array (FPGA). The devices may be implemented, in whole or in part, as a so-called application-specific integrated circuit (ASIC), i.e. an integrated circuit (IC) customized for their particular use. For example, the circuits may be implemented in CMOS, e.g., using a hardware description language such as Verilog, VHDL etc.
In an embodiment, pollution estimation system 100 comprises an acoustic sensor interface circuit, and an exhaust gas model circuit. The device 100 may comprise additional circuits, e.g., sensor circuit, audio compression circuit, a machine learning circuit. The circuits implement the corresponding units described herein. The circuits may be a processor circuit and storage circuit, the processor circuit executing instructions represented electronically in the storage circuits. The circuits may also be FPGA, ASIC or the like.
Many different ways of executing the method are possible, as will be apparent to a person skilled in the art. For example, the order of the steps can be varied or some steps may be executed in parallel. Moreover, in between steps other method steps may be inserted. The inserted steps may represent refinements of the method such as described herein, or may be unrelated to the method. For example, steps 430 and 440 may be executed, at least partially, in parallel. Moreover, a given step may not have finished completely before a next step is started.
A method according to the invention may be executed using software, which comprises instructions for causing a processor system to perform method 400. Software may only include those steps taken by a particular sub-entity of the system. The software may be stored in a suitable storage medium, such as a hard disk, a floppy, a memory, an optical disc, etc. The software may be sent as a signal along a wire, or wireless, or using a data network, e.g., the Internet. The software may be made available for download and/or for remote usage on a server. A method according to the invention may be executed using a bitstream arranged to configure programmable logic, e.g., a field-programmable gate array (FPGA), to perform the method.
It will be appreciated that the invention also extends to computer programs, particularly computer programs on or in a carrier, adapted for putting the invention into practice. The program may be in the form of source code, object code, a code intermediate source and object code such as partially compiled form, or in any other form suitable for use in the implementation of the method according to the invention. An embodiment relating to a computer program product comprises computer executable instructions corresponding to each of the processing steps of at least one of the methods set forth. These instructions may be subdivided into subroutines and/or be stored in one or more files that may be linked statically or dynamically. Another embodiment relating to a computer program product comprises computer executable instructions corresponding to each of the means of at least one of the systems and/or products set forth.
It should be noted that the above-mentioned embodiments illustrate rather than limit the invention, and that those skilled in the art will be able to design many alternative embodiments.
In the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. Use of the verb “comprise” and its conjugations does not exclude the presence of elements or steps other than those stated in a claim. The article “a” or “an” preceding an element does not exclude the presence of a plurality of such elements. The invention may be implemented by means of hardware comprising several distinct elements, and by means of a suitably programmed computer. In the device claim enumerating several means, several of these means may be embodied by one and the same item of hardware. The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage.
In the claims references in parentheses refer to reference signs in drawings of embodiments or to formulas of embodiments, thus increasing the intelligibility of the claim. These references shall not be construed as limiting the claim.
Number | Date | Country | Kind |
---|---|---|---|
16159645.7 | Mar 2016 | EP | regional |