System for locating a sound source, in particular noise pollution from vehicles
Patent Information
- Application Number
- EP2022838882
- Authority / Receiving Office
- EP · EP
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2021-12-31
- Filing Date
- 2022-12-22
- Publication Date
- 2024-11-06
Smart Images

Figure 1.1
Abstract
Description
SYSTEM FOR LOCATING A SOUND SOURCE, IN PARTICULAR SOUND POLLUTION FROM VEHICLES Field of invention
[0001] The present invention relates to the field of noise localization, in particular noise pollution. Noise pollution linked to transport, in particular that which affects residents living near traffic routes, constitutes a growing concern, most often associated with urban and peri-urban contexts. Diagnosing or predicting a noise environment requires knowledge of vehicle emissions in real traffic conditions.
[0002] The noise experienced is a source of stress and deterioration of health, causing numerous complaints from residents in urban or industrial areas, the objectivity and relevance of which should be verified.
[0003] To this end, various solutions have been developed to objectify the intensity, nature and location of mobile or fixed sound sources, in order to establish a map which can then be associated with topographic maps or photographic or video recordings, for example. Since noise is often geographically associated with transport and heavy industry, noise maps are also a source of pollution indices or environmental health problems.
[0004] The measurement is carried out by a sound level meter designed to measure the sound pressure level and provide a level expressed in decibels at a given point and time.
[0005] To enable localization of the sound source relative to the measurement point, a network of spatially offset microphones is generally used, and computer processing determines the direction of arrival of a sound wave by calculating its capture times by each microphone to deduce the delay (T) or the difference in arrival times (TDOA) recorded between two microphones. State of the art
[0006] Known in the state of the art is the article "Selection of the Closest Sound Source for Robot Auditory Attention in Moiti-source Scenarios" by Quang Nguyen and JongSuk Choi published in J Intell Robot Syst (2016) 83:239-251, describing a robot equipped with a sound source localization system. The robot's head has three microphones to localize sound sources in the azimuthal plane and an additional microphone attached to the chin of the robot's head to estimate the elevation. This first network is thus composed of four tetrahedron microphones that can estimate both the azimuth and the elevation of the sound source. The robot also has two microphones placed vertically on the base of the robot, to provide only the elevation.
[0007] Also known is the article “Passive acoustic system for tracking low-flying aircraft” by Alexander Sedunov, et al. published in IET Radar, Sonar & Navigation Volume 10, Issue 9 December 2016 Pages 1561-1568 describing an aircraft acoustic detection system designed for automated detection, classification, and tracking of low-flying aircraft using a network of passive acoustic sensors. The system consists of several autonomous sensor nodes, each equipped with a group of microphones, cameras, and electronic components that perform preprocessing and transmit the results wirelessly to a central processing station. The station fuses the sensor data to find the direction of arrival (DOA) of aircraft sounds and then uses triangulation techniques for target localization. The sensors are connected to an electronics box and a pan / tilt / zoom camera that is deployed on a separate tripod.The frame supports a microphone in the center, 240 cm from the floor, another directly above, 227 cm from the floor, and three more placed around the first to form an equilateral triangle with a radius of 122 cm, parallel to the floor.
[0008] Patent application US20170139031 describes an acoustic detection method comprising at least one group of acoustic sensors in communication with a computing device. The computing device is configured to process the received acoustic signals, and provide at least one detection of the presence of the acoustic source; determination of the direction of arrival of an acoustic wave emitted by an acoustic source; and classification of the acoustic source as to its nature. The cluster may comprise at least two sensors and the computing device may be configured to process the received acoustic signals and provide a location of the acoustic source in three dimensions.
[0009] US10261163 describes a system formed by at least one source, a plurality of sensors and a processing device. The source may emit one or more low-frequency sounds. The sensors may detect the low-frequency sounds and transform the low-frequency sounds into one or more signals comprising a plurality of data values. The processing device may operate in communication with the sensors to receive the signals and determine positioning information of the sensors based on the data values. Disadvantages of the prior art
[0010] The prior art solutions have various drawbacks.
[0011] The solution concerning the equipment of a humanoid robot for the detection of the origin of a sound source is suitable for operation at short distances (a few tens of centimeters) inside a building where stable atmospheric conditions prevail, with perfectly predictable and invariant sound propagation. Subjected to an outdoor environment, for precise localization at long distances, this solution would not be suitable because the sound propagation varies according to different atmospheric factors and outdoor use is disturbed by multiple and powerful parasitic noises.
[0012] Firstly, they are sensitive to parasitic noise caused by wind or rain as well as to thermal variations influencing the speed of sound propagation.
[0013] Second, the localization provided by prior art solutions is sometimes distorted when the sound source radiates diffusely or when two sound sources are close and compete with each other.
[0014] The solution proposed by Quang's paper is based on azimuthal localization of the microphone array. For each individual source, the elevations of the head and base arrays are estimated and triangulated to obtain the distance to the robot. Finally, the nearest source is determined, and its direction is used to control the robot. Accuracy is not very important; the goal is simply to discriminate the closest source among several sound sources. Solution provided by the invention
[0015] In order to overcome the drawbacks of the prior art, the present invention relates, in its most general sense, to a system for locating a sound source comprising an acoustic goniometer equipped with four microphones, an electrical signal processing circuit for calculating information representative of the direction of the sound source relative to said acoustic antenna, and an imaging means characterized in that it comprises two goniometers each equipped with a first pair of microphones on an axis XOX' and a second pair of microphones on a perpendicular axis YOY', said axes being perpendicular to the pointing direction of the goniometer, the four microphones of each goniometer being surrounded by an acoustically transparent shell,each of said goniometers further comprising an electronic circuit for processing the signals delivered by each microphone and the execution by a computer of an inter-correlation processing between said signals of the value of the time shifts of the signals supplied by the four microphones.,
[0016] Advantageously, the microphones arranged in pairs on two perpendicular axes are spaced apart by a distance of between 120 and 180 mm, and are arranged symmetrically with respect to the pointing axis.
[0017] Preferably, said microphones arranged in pairs on two perpendicular axes are spaced apart by a distance of between 100 and 500 mm, and are arranged symmetrically with respect to the pointing axis.
[0018] Preferably, the XOX' axis is offset along the pointing axis relative to the YOY' axis.
[0019] Preferably, said goniometers are vertically offset by a distance of between 50 and 500 cm and preferably between 200 and 500 cm.
[0020] According to a particular embodiment, said shell surrounding the four microphones of a goniometer is constituted by a frame covered with an acoustically transparent metal grid.
[0021] According to one variant, the system further comprises at least one thermometer for measuring the air temperature and providing said calculator with information on compensating for the speed of sound propagation.
[0022] Preferably, said cross-correlation processing is applied to each pair of microphones of an acoustic goniometer to determine the azimuth and elevation angles relative to the pointing direction.
[0023] According to another variant, the results of said inter-correlation processing are rejected in the case where the results obtained from a part of the pairs of microphones differ by a value exceeding a threshold value.
[0024] Advantageously, the angular orientation is calculated based on the average of the results obtained from the pairs of microphones.
[0025] According to a particular embodiment, the location of the sound source is determined based on an average value for the azimuth angle and by triangulation from the two elevation angles provided by the goniometers.
[0026] Detailed description of a non-limiting example of embodiment
[0027] The present invention will be better understood on reading the following description, concerning a non-limiting example of embodiment illustrated by the attached drawing where:
[0028] represents a perspective view of a system according to the invention. Description of an example of an implementation
[0029] Illustrates an example of embodiment of a system mounted on a mast (1) having the technical characteristics set out in claim 1.
[0030] It is composed of a first acoustic goniometer (100) and a wide-angle camera (300) and a second goniometer (200). The two acoustic goniometers (100, 200) and the wide-angle camera (300) are oriented in a pointing direction corresponding to an area of interest such that the optical axis of the camera (300) and the acoustic pickup axes of the two acoustic goniometers (100, 200) are substantially parallel, modulo the parallax, knowing that the distance from the area of interest is usually much greater than the distances between these axes.
[0031] The two goniometers (100, 200) are spaced about one hundred centimeters apart. They each include a set of four microphones respectively (110, 120, 130, 140) and (210, 220, 230, 240) grouped in pairs.
[0032] For the upper acoustic goniometer (100), the first pair of microphones (110, 120) is aligned on a substantially horizontal X1O1X1' axis, the microphones being arranged symmetrically with respect to the transverse plane of the system, and the second pair (130, 140) is aligned on a Y1O1Y1' axis perpendicular to the X1O1X1' axis and to the pointing direction, the microphones being arranged symmetrically with respect to the X1O1X1' axis. The four microphones are not necessarily arranged in the same plane. The X1O1X1' axis is, in the example described, offset in depth (i.e. along the pointing axis) with respect to the Y1O1Y1' axis.
[0033] Similarly, for the upper acoustic goniometer (200), the first pair of microphones (210, 220) is aligned on a substantially horizontal X2O2X2' axis, the microphones being arranged symmetrically with respect to the transverse plane of the system, and the second pair (230, 240) is aligned on a Y2O2Y2' axis perpendicular to the X2O2X2' axis and to the pointing direction, the microphones being arranged symmetrically with respect to the X2O2X2' axis. The four microphones are not necessarily arranged in the same plane. The X2O2X2' axis is in the example described offset in depth (i.e. along the pointing axis) with respect to the Y2O2Y2' axis.
[0034] In the example described, each acoustic goniometer (100, 200) consists of 4 microphones (110, 120, 130, 140) and (210, 220, 230, 240) positioned at the vertices of a tetrahedron with a side of 15 cm. The shape of the goniometer (100, 200) is determined to minimize the presence of material inside the volume of the tetrahedron in order to minimize acoustic interference (reflection, diffraction) that could harm the measurement or localization.
[0035] The four microphones (110, 120, 130, 140) and (210, 220, 230, 240) are positioned along parallel axes, pointed towards the road, along an axis of 35° to the horizontal. They are equipped with a protective windscreen (111, 121, 131, 141) and (211, 221, 231, 241) made of foam, for example 60mm in diameter for wind and rain protection.
[0036] Since the device is used outdoors, the speed of sound is highly dependent on temperature (at an outdoor temperature of -20°C, sound travels at a speed of 319m / s, while at an outdoor temperature of +50°C, sound travels at a speed of 361m / s). The speed of sound is also somewhat dependent on humidity.
[0037] To ensure sufficient accuracy for the device to be used for metrology purposes, the device includes an external temperature sensor and preferably an external humidity sensor providing signals that allow a computer to correct for differences in arrival times between the different microphones and angles. The code executed by the computer takes into account the signals generated by the microphones as well as the data provided by these temperature and humidity sensors, generating additional input parameters.
[0038] The entire goniometer (100, 200) is placed in a vandal-proof protective cage (105, 205). This cage (105, 205) is made of an acoustically transparent material.
[0039] An alternative is to replace the four windshields (111, 121, 131, 141) and (211, 221, 231, 241) with a single foam block occupying the entire volume of the cage (105, 205). The advantage is to slightly increase the insensitivity to wind and rain, at the expense of the isotropic character presented by a spherical windshield.
[0040] All measurements and location calculations are performed on an electronic card embedded in the goniometer (100, 200). The calculation code is embedded in a non-volatile memory installed on a microcontroller. The results are sent to the central unit in the form of UDP frames.
[0041] The microphones (110, 120, 130, 140) and (210, 220, 230, 240) are, for example, analog MEMS microphones. They have their own internal differential output preamplifier, connected to a 4-channel 24-bit analog-to-digital converter (ADC) via DC filter capacitors. They have an internal temperature and humidity sensor as well as a heating resistor to ensure that the microphone temperature remains above a minimum value (e.g. 10°C).
[0042] They have a background noise level of 20 dBA and an acoustic saturation level of 123 dB. On the frequency response spectrum, the -3dB points are located at 20 Hz and above 20 kHz.
[0043] The microphones are in the classic ½ inch format, allowing the use of multi-frequency calibrators and controllers on the market and have their own internal eeprom type memory in which calibration information can be recorded and tracked (microphone serial number, date of commissioning, sensitivity in mV / Pa, possible frequency corrections, dematerialized metrological logbook, etc.).
[0044] The analog / digital converter is also calibrated to ensure the metrological nature of the analog-to-digital conversion. An eeprom is also used to store the calibration parameters for each of the 4 measurement channels.
[0045] The microphone is connected to the goniometer using a snap-in connector. The microphones (110, 120, 130, 140) and (210, 220, 230, 240) are therefore field replaceable and can be used as is without additional calibration, although it is entirely possible to require verification or calibration in the field in the operating procedures.
[0046] For the initial calibration or laboratory verification phases, it is possible to inject a differential signal in place of each microphone, taking the precaution of connecting the ground of the goniometer to the ground of the control devices. Furthermore, it is quite possible to inject simulated signals on the 4 channels of the goniometer by means of a synchronous 4-channel DAC card in order to qualify the complete operation of the goniometer (measurement and location) using standard signals.
[0047] Optionally, acoustic transmitters are arranged inside the cage (105, 205) of the goniometers (100, 200). The purpose of this acoustic transmitter device is to carry out periodic self-tests to validate the operation of the location function (with pink noise) and the absence of measurement drift (pink noise and / or sinus).
[0048] Principle of localization of the dominant acoustic source
[0049] Each of the acoustic goniometers (100, 200) makes it possible to determine the direction of origin of a dominant noise, in the form of two angles: the azimuth (angle in the horizontal plane) and the elevation (angle in the vertical plane relative to the pointing axis). The combination of the angular information makes it possible to determine the absolute position of the detected sound source.
[0050] The technique used consists of determining by inter-correlation between the signals the value of the time shifts observed by the four microphones of each of the acoustic goniometers (100, 200), then finding to which direction of origin this corresponds.
[0051] The measured time lag is the result of the propagation of sound between two microphones (110, 120, 130, 140) separated by a fixed distance (15cm), it is necessary to have an indication of the outside temperature in order to be able to compensate for the variation in the speed of sound as a function of the temperature. An accuracy of the order of 2 or 3°C on the temperature measurement is sufficient. Three temperature measurement sensors are installed on the mast (1) in order to make the determination of the outside temperature more robust.
[0052] However, to improve accuracy, a variation is to equip the device with a hygrometer, to take into account the degree of humidity of the air in addition to its temperature.
[0053] The operation of processing time shifts to determine the azimuth and elevation angles takes advantage of the redundancy of the equations to be solved (6 time shifts corresponding to 6 pairs of microphones, therefore 6 equations for 2 unknowns, the azimuth and elevation angles).
[0054] These treatments make it possible to determine a focusing indicator on which a threshold is applied to determine whether the calculation (azimuth, elevation) obtained is reliable.
[0055] The signal is considered reliable when all the signals from the four microphones of an acoustic goniometer (100, 200) are coherent and focus on a single direction. This situation may not occur punctually when two sources located in different directions are in competition, or more frequently when a source does not have a sufficiently punctual character because it "radiates" diffusely. In the case where the test of this indicator is negative, and a treatment eliminates the measurement point.
[0056] This operating characteristic is important since it is possible to eliminate certain locations that do not present sufficient guarantees of focusing, which is particularly interesting in legal metrology.
[0057] A single system cannot correctly determine the distance to the source. To do this, the microphones would have to be unreasonably far apart.
[0058] For this reason, the invention proposes to combine two goniometers (100, 200) of limited size (15cm), installed vertically 1 meter from each other, and to carry out a triangulation treatment to determine the location of the source.
[0059] In this method, the azimuth angles obtained by the two goniometers (100, 200) are compared. If they are too far apart, the goniometers (100, 200) are considered to have focused on different sources and the measuring point is eliminated. If they are close enough, the location is determined based on an average value for the azimuth angle and by triangulation from the two elevation angles to determine the source point.
[0060] It is then possible to calculate the distance to the source for each microphone, then to recalibrate the measured sound levels to a reference distance D ref (7.6 meters, the value proposed because it is commonly used in track tests). It is therefore on the basis of these levels adjusted to a standard distance that a level overshoot is noted.
[0061] The formula we use for the recalibration is a purely geometric compensation in 10 log10 (d / d ref ) considering that the distances to the source remain small (<30 m) and that we can therefore neglect the variations in atmospheric absorption as a function of frequency.
[0062] Evaluation of uncertainties related to the localization technique
[0063] By proceeding by crossing two directions, the distance to the source obtained by calculation is necessarily different from the actual distance. It can be either smaller, generating an error that could be to the advantage of the offender, or larger, generating an error that could be unfavorable to the offender, which is unacceptable in such a system.
[0064] Sound level measurements – warning about time weighting
[0065] Each goniometer (100, 200) produces for each of the four microphones (110, 120, 130, 140) and (210, 220, 230, 240) an equivalent overall sound level weighted A and C at the desired periodicity.
[0066] To assess the position of the dominant sound source and the level emitted by it, the system divides the acoustic signal into time intervals, for example between one tenth of a second and one fiftieth of a second.
[0067] Over a given time interval, a sound level Leq is calculated in a given direction. However, for the same time interval, the calculation of LFmax can be strongly impacted by a loud noise emitted by another source a few time intervals earlier, such as a detonation at the exhaust pipe, a horn blast, or the presence of a jackhammer, etc.
[0068] A high LFmax value over a time interval will continue to have a major impact on the level for a number of consecutive time intervals, with a decrease of approximately 3.5 dB per tenth of a second. During subsequent time intervals, it is quite possible that the focus will be on a source other than the one initially responsible for the measured level.
[0069] It is therefore impossible to assert that the Fast level measured during a time interval is solely attributable to the source considered dominant during this time interval.
[0070] It is therefore preferable that the level retained be obtained from the list of equivalent levels that can be assigned with certainty to the same source, the list itself having to represent a certain minimum duration in order to guarantee that the same source could be observed for a time estimated as sufficient (for example: 0.5 sec). Assigning a measuring point to a vehicle
[0071] The system according to the invention provides a set of measurements that must be linked to the correct vehicle. In order for an infraction to be detected, it is preferable for the measurement to be carried out over an entire area upstream and downstream of the device, and not just in line with the post, because it would probably be too easy to cut and then re-start the throttle once the position of the radar is known. For this reason, the direction of the dominant acoustic source is linked to the correct vehicle by carrying out a video analysis of the scene over the entire passage of the vehicle.
[0072] In the event of a potential violation (i.e. in the case where a source located in the detection zone emits a level higher than the defined limit, this level being understood as rectified at the reference distance, and reduces measurement and method uncertainties), the video images are analyzed in several phases:
[0073] 1-Use of an image-by-image “boxing” algorithm to detect each vehicle, frame it with a rectangle, and classify it by family (VL, PL, 2R, at least)
[0074] 2-Elimination of measuring points that cannot be linked with certainty to a single vehicle
[0075] 3-Reconstruction of the trajectory of the vehicles present in the scene (chaining of boxes) and verification of the verbalization criteria (exceeding the level for a sufficient time on the trajectory of the same vehicle)
[0076] 4-Matching the trajectory of the offending vehicle with the LAPI data in order to find the identification of the vehicle.
[0077] In phase 2, the criterion for assigning a measurement point to a vehicle is that the point representing the acoustic direction is located within the rectangle representing the vehicle. When a point is not located within any rectangle, the measurement is eliminated. When a point is located within more than one rectangle, this means that it can potentially be assigned to more than one vehicle. It is therefore also eliminated. Additional criteria may be added during real-life testing to ensure that a measurement point can be assigned to a vehicle with certainty.
[0078] In phase 4, the LAPI subsystem (which operates autonomously and permanently) will be requested by providing a time interval determined from the movement of the vehicle in the 180° scene captured by the fish-eye camera, which covers the LAPI detection zone. Composition of the infringement message
[0079] For an application concerning the verbalization of the authors of noise pollution, the information calculated by the aforementioned system is transmitted as metadata within the framework of the offense message, respecting the required encryption constraints.
[0080] Regarding the photos, the system can provide several photos in the violation message: A 180° photo of the moment when the noise level attributable to the vehicle is at its maximum. This photo will allow the entire scene to be checked and to ensure that there is indeed a single vehicle in the identified acoustic direction, and no other vehicle in the immediate vicinity. A second 180° photo of the moment when the plate was read by the ANPR device, playing the usual role of context image. A third detailed photo, from the ANPR device, allowing the plate read to be checked, if necessary. This photo may possibly be embedded in the second photo if the authority imposes a maximum of two images.
[0081] An identically colored frame will be placed on both 180° photos to materialize the fact that the vehicle's trajectory was detected and that it was possible to link the vehicle's position in the two 180° images.
[0082] Angular matching of camera and acoustic goniometers
[0083] The proper functioning of the system relies on the ability to accurately match the acoustically calculated azimuth and elevation angles with the images from the wide-angle fisheye camera (300).
[0084] Concerning the acoustic determination of angles, the angular precision obtained during the operation of the system can be calculated by numerical simulation (this is less than 1.14° at 102.4 kHz and 0.72° at 204.8 kHz). By construction, the position of the four microphones in space is guaranteed with a largely sufficient precision, and the uncertainty actually depends on only two factors: the estimation of the speed of sound (which will be compensated for temperature and humidity) and the sampling frequency of the acoustic signals. It is this last factor which mainly conditions the angular precision, for a given inter-microphone distance (15cm).
[0085] Obviously, doubling the dimension of the tetrahedron has the same effect as doubling the sampling rate.
[0086] Concerning the projection of angles in the image, this is based on the fact that the fish-eye camera (300) provides a 180° image according to a so-called "equirectangular" projection. This type of projection is equivalent to the concept of a planisphere, and allows a direction (azimuth x elevation) to be directly matched with a point in the image.
[0087] The last point concerns the relative positioning of the two acoustic goniometers (100, 200) and the fish-eye camera (300), which must obviously be mechanically guaranteed, both in terms of the position and the orientation of each of the elements. This point also presents no difficulty from the moment when all the elements are mounted on a common plate.
[0088] Observation of a vehicle within a complex scene
[0089] While it is easy to make an observation when only one vehicle is present in the scene, it becomes more complicated when several vehicles cross, follow each other closely or drive alongside each other.
[0090] In this type of situation, it is impossible to do without video analysis to rule out possible cases of confusion. Analyzing trajectories and matching acoustic directions with the position of all vehicles is essential to ultimately retain only the non-disputed measurement points. Road noise treatment
[0091] The noise level measured when a vehicle passes over it includes the noise caused by rolling on the road surface. This noise is significantly affected by wet roads, and it is recommended to correct or even invalidate the measurements when the road surface is not dry.
[0092] To enable the device to take into account the condition of the road, the invention provides for the use of a contactless road condition detector, an optical sensor in the infrared spectrum, directed towards the road which is capable of indicating whether the road is dry, wet, soaked, ice or a water / ice mixture.
[0093] The system performs a real data fusion between the wide-angle imaging part and the acoustic part in order to exclude measurements which are not related to the vehicle: the device according to the invention follows the vehicle in the scene and the acoustic angles are compared with the viewing angles of the different vehicles.
Claims
- A sound source localization system for characterizing transport-related noise pollution, comprising an acoustic goniometer equipped with four microphones (110, 120, 130, 140) and (210, 220, 230, 240), an electrical signal processing circuit for calculating information representative of the direction of the sound source relative to said acoustic goniometer, and an imaging means (300) characterized in that it comprises a second goniometer (200), said goniometers (100, 200) being equipped with a first pair of microphones (110, 120) and (210, 220) on an axis XOX' and a second pair of microphones (130, 140) and (230, 240) on a perpendicular axis YOY', said axes being perpendicular to the pointing direction of the goniometer, the four microphones (110, 120, 130, 140) and (210, 220, 230, 240) of each goniometer (100, 200) being surrounded by an acoustically transparent shell (105, 205), each of said goniometers (100,200) further comprising an electronic circuit for processing the signals delivered by each microphone (110, 120, 130, 140) and (210, 220, 230, 240) and the execution by a computer of an inter-correlation processing between said signals of the value of the time offsets of the signals provided by the four microphones (110, 120, 130, 140) and (210, 220, 230, 240) and in that the system further comprises at least one thermometer for measuring the temperature of the outside air and / or a hygrometer for measuring the humidity of the outside air, and providing said computer with information for compensating the speed of sound propagation. - Sound source localization system according to claim 1 characterized in that the microphones (110, 120) and (210, 220) arranged in pairs on two perpendicular axes are separated by a distance of between 50 and 1000 mm, and are arranged symmetrically with respect to the pointing axis. - Sound source localization system according to claim 1 or 2 characterized in that the XOX' axis is offset along the pointing axis relative to the YOY' axis. - Sound source localization system according to claim 1 characterized in that said goniometers (100, 200) are vertically offset by a distance between 50 and 500 cm. - Sound source localization system according to claim 1 characterized in that said shell (105, 205) surrounding the four microphones of a goniometer is made up of a frame covered with an acoustically transparent metal grid. - A sound source localization system according to claim 1 characterized in that said cross-correlation processing is applied to each pair of microphones (110, 120, 130, 140) and (210, 220, 230, 240) of an acoustic goniometer to determine the azimuth and elevation angles relative to the pointing direction. - Sound source localization system according to the preceding claim characterized in that the results of said inter-correlation treatments are rejected in the case where the results obtained from a part of the pairs of microphones (110, 120, 130, 140) and (210, 220, 230, 240) differ by a value exceeding a threshold value. - Sound source localization system according to claim 6 and / or 7 characterized in that the angular azimuth orientation is calculated as a function of the average of the results obtained from the pairs of microphones (110, 120, 130, 140) and (210, 220, 230, 240). - Sound source localization system according to claim 6 characterized in that the localization of the sound source is determined as a function of an average value for the azimuth angle and by triangulation from the two elevation angles provided by the goniometers (100, 200). - Sound source localization system according to claim 1 characterized in that it further comprises a non-contact road condition detector.
