Detecting, locating and tracking human postures by means of coherent radar

Abstract

The invention relates to a method for detecting the posture of an individual, the method comprising the following steps: [100] continuously transmitting coherent signals; [200] continuously collecting the echoes of the signals transmitted in step [100] and extracting the corresponding beat signals; [300] extracting, from the beat signals, the range-Doppler signals; [400] assigning a 3D position to each potential target; [500] determining whether the potential target corresponds to the presence of an individual and assigning a state thereto, where applicable, characterised in that step [500] comprises a step [520] consisting in using a statistical test to compare the distribution of the potential targets with a reference distribution.

Description

Detection, localization and tracking of human postures by a coherent radar. Technical field of the invention The present invention belongs to the field of personal safety. More particularly, the present invention relates to a method for identifying the posture of an individual and detecting possible falls. State of the art With an aging population, the safety of older adults living at home or in institutions has become a key concern. Among the problems these individuals face, falls are a significant cause of death when third parties cannot quickly intervene to provide first aid. Devices with alarms in the form of watches and push-button collars are already available, allowing a person immobilized on the ground to sound the alarm. However, the effectiveness of these devices depends on the wearer. The wearer must be conscious to trigger the alarm. Other devices worn by the person to be monitored can also be triggered automatically when a sudden change in posture or physiological parameters is recorded. However, the effectiveness of such devices is always dependent on the user, who must agree to wear them throughout the day. It is often observed that the user forgets to put the device back on after changing clothes. Furthermore, these latter devices are often considered too intrusive and pose data confidentiality problems. The use of radar for tracking people at home has been proposed. The radar signal does not pose privacy issues, allowing the sensor to be installed in the bathroom or bedroom. However, although the use of radar is well known to those skilled in the art, the interpretation of the results obtained is particularly difficult. A poorly discriminating interpretation will result in a significant number of false positives. As a result, those responsible for monitoring will gradually stop paying attention to the alerts received. Alternatively, an overly drastic interpretation of the data collected risks excluding important events that could have caused the death of the person being monitored. It is therefore necessary to implement a data processing method and a device that allows: - To accurately discriminate between one and preferably several individuals in the same scene. - To continuously determine the position of each of these individuals. - To accurately detect any change in posture of the individuals being monitored. - To monitor the physiological parameters of said individuals. Thus, the present invention relates to a method for detecting the posture of an individual comprising the steps:

[0100] continuously emitting coherent signals

[0200] continuously collecting echoes of the signals emitted in step

[0100] and extracting the corresponding beat signals,

[0300] extracting from said beat signals the range-Doppler signals,

[0400] assigning each potential target a 3D position,

[0500] determining whether the potential target corresponds to the presence of an individual and assigning it a state if so,characterized in that step

[0500] comprises a step

[0520] consisting of using a statistical test to compare the distribution of the potential targets to a reference distribution. In the context of the present invention, the term "coherent" is intended to designate signals emitted by a radar in which the phase of the manipulated signals (in transmission and reception) is controlled, which thus makes it possible to recover information on the Doppler frequency of the target. Preferably, said coherent signal is of pulsed Doppler type waveform, phase encoded or FMCW (frequency modulated continuous waves). Most preferably, said coherent signal is emitted by an FMCW type radar. For the purposes of the present invention, the term "target" refers to any moving object returning an echo whose beat signal amplitude is greater than a predefined threshold. According to one embodiment of the invention, the steps mentioned above are implemented successively in the order in which they are described. Alternatively according to another preferred embodiment of the invention, the mentioned steps can be implemented in a different order than that described. According to a preferred embodiment of the invention, step

[0500] further comprises the step:

[0530] grouping a plurality of 3D positions corresponding to an individual and analyzing the geometric and statistical distribution of the plurality of 3D positions and assigning a posture to the latter. According to a preferred embodiment of the invention, step

[0530] further comprises an analysis of the Doppler signals associated with said plurality of 3D positions. According to a preferred embodiment of the invention, said posture assigned to step

[0530] is chosen from the group consisting of “static standing”, “walking standing”, “sitting” and “lying down”. According to a preferred embodiment of the invention, the method further comprises the step:

[0540] comparing the successive postures attributed in step

[0530] and attributing a state to said individual. According to a preferred embodiment of the invention, said step

[0540] is carried out by measuring the Mahalanobis distance. According to a preferred embodiment of the invention, said state assigned to step

[0540] is chosen from the group consisting of “static”, “movement”, “fall” and “not fall”. According to a preferred embodiment of the invention, the statistical test used in step

[0520] is a Kolmogorov-Smirnov test. According to a preferred embodiment of the invention, step

[0300] is carried out by distance processing

[0310] then by Doppler processing

[0320] . According to a preferred embodiment of the invention, step

[0400] comprises:

[0410] a step of extracting the distances of said potential targets, and

[0420] a step of extracting the elevation and azimuth angles of said potential targets. According to a preferred embodiment of the invention, the method further comprises a step:

[0600] measuring the respiratory and cardiac parameters of said individual. According to a preferred embodiment of the invention, step

[0600] is carried out from the Doppler signals associated with said individual. The present invention also relates to a device for implementing a method according to the invention, characterized in that it comprises a coherent radar and a computer programmed so as to carry out the steps of the method according to the invention. According to a preferred embodiment of the invention, said coherent radar is fixed at a height of between 150 and 250 cm in an adjustable manner. Said radar can be indifferently associated with a vertical structure (e.g. wall, post) or horizontal structure (e.g. ceiling). According to a preferred embodiment of the invention, said coherent radar operates at a wavelength suitable for the detection of human tissues and with a bandwidth allowing a spatial resolution of less than 1 meter. According to a preferred embodiment of the invention, said coherent radar comprises at least one transmitting antenna and a set of receiving antennas, grouped into two networks distributed into two groups organized along two perpendicular axes. Brief description of the figures general architecture of an embodiment of a device according to the invention. Synopsis of the method according to the invention. Synopsis of the classification method according to the invention. Detailed description of the invention With reference to, and are described below - a preferred embodiment of the invention.

[0100] continuously emit coherent signals The method according to the invention is based on the analysis of echoes of coherent radar signals, for example an FMCW (i.e. frequency modulated continuous wave) radar. This latter type of radar is well known to those skilled in the art and commercially available. An FMCW radar is characterized by the continuous emission of frequency modulated waves. It makes it possible to determine both the distance and the speed of movement of a target, if applicable. The determination of the direction (angles) is possible thanks to antenna diversity in reception. The radar used in the method according to the invention is placed in the enclosure intended to be monitored. This enclosure is, for example, a bedroom or living room of an elderly person likely to fall. However, the method according to the invention is not limited to the monitoring of a particular public and can be used for all types of people in all types of environment. The radar used in the method according to the invention is preferably associated with a data processing device (i.e. calculator), making it possible to collect and process the data from said radar. Said data processing device may be located near said radar in the enclosure intended to be monitored. Alternatively, said data processing device may be remote from said radar and connected to the latter by a computer link (e.g. internet). Said data processing device is programmed so as to be able to implement the algorithm described below. The choice of the computer, its configuration and the programming language used to implement the method according to the invention is within the reach of the person skilled in the art and does not present any particular difficulties. According to a preferred embodiment of the invention, said data processing device is associated with an alarm device making it possible to warn a third party in the event of detection of a particular event (e.g. fall). This alarm device can take different forms, among which we can notably cite light warning devices, audible warning devices, automated telephone calls, sending of messages (e.g. SMS, e-mail). Continuous wave radars continuously transmit a microwave signal. The received echo is therefore continuously processed. The use of a frequency modulated wave radar allows, by varying the emitted signal, to measure the frequency shift between the emitted and received signals and thus to work out the radar-target distance. Preferably, the radar used in the method according to the invention contains a transmitting antenna, 4 receiving antennas uniformly distributed along an x ​​axis and 4 other receiving antennas, arranged equidistant along a z axis perpendicular to the x axis. The two receiving antenna arrays allow angular detection in azimuth and elevation. Using a frequency modulated wave, beat signals between the emitted and returned signals can be obtained and the distance to a target can be measured. Preferably, the radar used in the method according to the invention operates at 24.15 GHz with a bandwidth of 200 MHz.

[0200] continuously collecting the echoes of the signals transmitted in step

[0100] and extracting the corresponding beat signals. According to a preferred embodiment of the invention, step

[0200] comprises: -

[0210] continuously collecting the echoes of the signals emitted in step

[0100] , -

[0220] extracting the beat signals which result therefrom. The radar used in the method according to the invention emits a continuous signal which is reflected by the different surfaces present in the analyzed scene. For example, in the method according to the invention, the signal emitted in step

[0100] will be reflected by the walls of the room, the furniture and any people present in the monitored area. In step

[0210] , the reflected signals (i.e. echoes) are collected by the receiving antennas. The radar transmits a continuous wave modulated in frequency over a period of time. The reflected signal has the same shape as the transmitted signal, but it is shifted in time, because the round trip between the radar and the target takes a time which is proportional to the distance of the target. If the reflected signal is monitored over several periods, an additional frequency shift will be detected for a target approaching or moving away from the radar, due to the Doppler effect. Finally, since the method according to the invention uses several antennas distributed in space, the direction of arrival of the signal can be established, in order to obtain the 3D position of the target. This collection of signals in step

[0210] is carried out on each of the reception antennas. In the preferred embodiment presented here, the radar used will therefore collect eight different series of data per transmitted signal.

[0220] The beat signals are extracted from the collected signals and the transmitted signal. These beat signals will then undergo a certain number of processing operations to determine in particular the position and speed of each of the echoes. Indeed, to obtain a complete detection, spatio-temporal processing of the beat signal is necessary. For this, the signal is preferably digitized and saved in the form of a 3D radar cube for each antenna array for further processing.

[0300] extracting from said beat signals the distance-Doppler signals. According to a preferred embodiment of the invention, step

[0300] comprises the steps:

[0310] mathematical processing of the beat signals to obtain the range-time signals, and

[0320] Doppler processing of the beat signals. Among the mathematical treatments that can be used within the framework of the method according to the invention, we can preferentially cite the Fourier transform and even more preferentially the fast Fourier transform. The Fourier transform was developed by French mathematician Jean Baptiste Joseph Fourier. The Fourier transform has become a fundamental method in signal processing procedures because radar echoes contain a variety of information. In this case, thanks to fast Fourier analysis, useful information (distance, Doppler) can be extracted from the complex signals extracted in step

[0220] . Discrimination between "moving echoes" and "stationary echoes" is a technology well known to those skilled in the art of radar echo processing. Stationary targets return a radar echo at the frequency of the received signal, while moving targets change the frequency of the returned signal. This effect is known as the Doppler effect. The difference between the received frequency (as transmitted by the radar) and the returned frequency (as perceived by the radar) is a function of the target's velocity relative to the radar. Step

[0320] of the method according to the invention will therefore preferably consist of Doppler processing and even more preferably of VCM (Moving Target Visualization, in English MTI [1] for “moving target indicator”) type processing, of “coherent phase difference” (CPD) type [2] or of “pulse-doppler processing” (PDP) type [3]. VCM processing is well known to those skilled in the art. In short, it involves simple high-pass filtering of time-distance signals. Indeed, stationary objects do not change the frequency of the transmitted signal, but moving objects do. Consequently, high-pass filtering eliminates signals from stationary objects. CPD consists of subtracting coherently for each distance the samples of the distance-time signals from that corresponding to the stationary clutter, obtained by averaging the different samples over time for the same distance. PDP is a fast Fourier analysis of the distance-time samples for each distance. It allows obtaining the range-Doppler signals. Filtering is then necessary to isolate the clutter and extract the Doppler signals corresponding to the moving targets.

[0400] assign a 3D position to each potential target. According to a preferred embodiment of the invention, step

[0400] comprises: - a step

[0410] of detecting (in distance) the potential targets by CFAR processing [4] which makes it possible to determine the distances of said potential targets from the signals obtained at the end of step

[0300] - an angular processing step

[0420] for each distance determined in step

[0410] to obtain a 3D position. Step

[0400] can be carried out in particular by determining on the one hand the angular position (elevation and azimuth angles) of the potential target relative to the radar and the distance of the potential target relative to the radar on the other hand. The combination of these different data makes it possible to precisely position the potential target relative to the radar. The determination of the angular position of each potential target can be made by different techniques known to those skilled in the art. According to a preferred embodiment of the invention, this determination is made by the Multiple Signal Classification (MUSIC) algorithm [6]. Other algorithms such as Capon [7], ESPRIT [8], etc. can be used as required. At the end of steps

[0100] -

[0400] , the method according to the invention was able to isolate and position a certain number of moving targets. The following steps of the method according to the invention aim to determine whether these targets correspond to an individual and, if necessary, to determine the posture of said individual and finally to detect sudden changes in posture of said individual. On the basis of these last steps, alerts can be sent to signal the fall of the individual.

[0500] determine whether the potential target corresponds to the presence of an individual and assign it a status if so With reference to the present invention, a classification process preferably carrying out step

[0500] of a method according to the invention is presented. This is a four-step process that first confirms the presence of the detected person, then determines their posture (static standing, walking standing, sitting, lying down) and finally deduces their state (fall or no fall). According to a preferred embodiment of the invention, the method further comprises the step:

[0510] filtering spatial and / or Doppler signals from the Doppler processing in step

[0320] , Preferably, the first step of the classification process consists of carrying out spatial filtering, for example by the beamforming technique [9] to consider only the signals coming from the distance-angle zone where the target is detected. Spatial filtering is optional. Depending on the purpose of the processing, Doppler filtering can also be considered to consider only the signals belonging to a certain Doppler range, or even a double Spatial and Doppler filtering.

[0520] consisting of using a statistical test to compare the distribution of potential targets to a reference distribution. Signals corresponding to potential targets are subjected to a statistical test to determine whether their distributions correspond to the expected distribution of an individual. Preferably, said statistical test is a Kolmogorov-Smirnov test, to confirm or deny the presence of an individual. The experiment conducted by the applicants showed a difference in the distribution characteristics of the signals in the presence and absence of a human target. Other types of tests can be used. The Kolmogorov-Smirnov test is used to determine whether two samples follow the same distribution. Therefore, it can be easily used to determine whether a plurality of samples of a signal correspond to the presence of an individual by comparing their distribution to a reference distribution. The main purpose of these steps is to determine whether the "bright spots" (or "targets") identified by the previous treatments (

[0510] ) correspond to humans or to objects in the room that have variable reflectivity (SER). It is well known to those skilled in the art that each type of target corresponds to a specific SER distribution; this technique is commonly used for classification. For example, "Swerling" type distributions (I, II, III, etc.) are classics. Here, we perform a test of conformity to a "reference" distribution constructed from a model produced experimentally. If the difference between the two distributions is below a predefined threshold, then it is decided that the said signals correspond to those of an individual. Reference distributions can be easily obtained by collecting a large amount of data from tests involving an individual in different configurations. The resulting database can be stored and used in the information processing device associated with the radar. The method according to the invention is continued if the result of the statistical test is positive, otherwise the presence of an individual is rejected and the method according to the invention is repeated at steps

[0100] -

[0400] . Due to its robustness, the central use of the statistical test in the method according to the invention makes it possible to improve the quality of the results obtained compared to other methods in the state of the art even when the signal / interference ratio is unfavorable.

[0530] grouping a plurality of 3D positions corresponding to an individual and analyzing the geometric and statistical distribution of the plurality of 3D positions and assigning a posture to the latter. When the detection of an individual is confirmed by the statistical test, the different associated 3D positions are subjected to a grouping procedure ("Clustering"). The aim of this step is to group the different 3D points into groups of points, each describing an individual. This makes it possible to exclude irrelevant points and those that would come from errors. This technique, known as data clustering, aims to divide a set of data into different homogeneous groups, in the sense that the data in each subset share common characteristics, which most often correspond to proximity criteria. In the context of the present invention, step

[0530] preferentially implements a method chosen from the group consisting of DBSCAN (Density-Based Spatial Clustering) [5], LDA (Linear Discriminant Analysis) [6] or k-means [7]. At the end of the clustering step we therefore obtain a set of points located in the space associated with a single individual. The set of points grouped in the previous step has a distribution in space that depends directly on the posture of the individual. For example, a standing individual will correspond to a distribution of the cluster points in a cylinder / ellipsoid with a vertical axis of revolution, whereas a lying individual will correspond to a distribution of the cluster points in a cylinder / ellipsoid with a horizontal axis of revolution. Advantageously, only the bright points which have a "human" fluctuation are kept, making it possible to reduce their number. The objective then is to determine what the posture of the person is. The approach consists of finding, among a finite and predetermined set of postures, what is the most probable posture by making: - the hypothesis of a particular posture, for example "standing" for a simplified "digital twin" making it possible to characterize the position of the different members and their orientation, - for each bright point from

[0520] , we estimate the probability that this point is indeed from this particular posture (calculation of "likelihood" seeking to estimate to what extent this target being analyzed can be from the posture considered) - after evaluation of the "likelihood" of all the bright points with the posture under test, assignment of an "overall" probability. This operation is performed for each of the possible postures. The most probable posture is then retained. In doing so, it is possible to assign a posture chosen from the group consisting of "static standing", "walking standing", "sitting" and "lying down".

[0540] comparing the successive postures assigned in step

[0530] and assigning a state to said individual. The main purpose of this step is to determine whether the monitored individual is in a state requiring the triggering of an alarm. This is particularly the case when the monitored individual falls. This latter state is associated with a sudden change in the spatial distribution of signals associated with an individual. This step can be implemented by comparing the successive characteristics of the signals corresponding to an individual from step

[0510] and / or directly the evolution of the posture determined in step (530]. The result of the clustering is combined with a statistical evaluation to determine the state (fall or no fall) of the person. The statistical evaluation is preferably based on a statistical comparison of the characteristics obtained in step

[0510] with respect to the characteristics of the same target at the preceding measurement instant. Thus a change in state of the target between two measurement instants will produce a difference in characteristics according to the states of the target at the two instants. The statistical test used at this stage is preferably the evaluation of the Mahalanobis distance. Depending on the result of the statistical test used, the state assigned to step

[0540] is chosen from the group consisting of “static”, “movement”, “fall” and “no fall”. Advantageously, the method according to the invention may further comprise a step

[0600] consisting of measuring the respiratory and cardiac parameters of the detected individual. This step

[0600] may in particular be implemented by analyzing the Doppler signals associated with said person. These methods are well known to those skilled in the art. The measurement of these parameters can in particular be done from Doppler signals, using correlation or the so-called cepstrum method [8] or by the method described by Chioukh [9]. Step

[0600] is preferably carried out when the posture, attributed to the individual detected, in step

[0520] is chosen from the group consisting of “static standing” and “lying down”. References [1] MI Skolnik, Introduction to Radar Systems, 2nd ed. McGraw Hill, Inc., 1980. [2] E. Hyun, Y.S. Jin, And J.H. Lee, A pedestrian detection scheme using a coherent phase difference method based on 2D range-Doppler FMCW radar. Sensors, 2016, vol. 16, no 1, p. 124. [3] M. A. Richards, J. A. Scheer, W. A. Holm, Principles of Modern Radar - Vol. I : Basic Principles, SciTech Publishing Inc, 2010. [4] H. Rohling, "Radar CFAR Thresholding in Clutter and Multiple Target Situations," in IEEE Transactions on Aerospace and Electronic Systems, vol. AES-19, no. 4, pp. 608-621, July 1983, doi: 10.1109 / TAES.1983.309350. [5] P. S. Diao, T. Alves, B. Poussot and S. Azarian, "A review of Radar Detection Fundamentals," in IEEE Aerospace and Electronic Systems Magazine, doi: 10.1109 / MAES.2022.3177395. [6] B. Li, S. Wang, J. Zhang, X. Cao, & C. Zhao, "Fast-MUSIC for automotive massive-MIMOradar". arXiv preprint arXiv:1911.07434, 2019, pp. 1-15. [7] C. D. Richmond, "The CAPON-MVDR algorithm: threshold SNR prediction and the probability of resolution," 2004 IEEE International Conference on Acoustics, Speech, and Signal Processing, Montreal, QC, Canada, 2004, pp. ii-217, doi: 10.1109 / ICASSP.2004.1326233. [8] S. Kim, D. Oh and J. Lee, "Joint DFT-ESPRIT Estimation for TOA and DOA in Vehicle FMCW Radars," in IEEE Antennas and Wireless Propagation Letters, vol. 14, pp. 1710-1713, 2015, doi: 10.1109 / LAWP.2015.2420579. [9] B. D. Van Veen and K. M. Buckley, "Beamforming: a versatile approach to spatial filtering," in IEEE ASSP Magazine, vol. 5, no. 2, pp. 4-24, April 1988, doi: 10.1109 / 53.665

[0010] Z. Yu et al., "A Radar-Based Human Activity Recognition Using a Novel 3-D Point Cloud Classifier," in IEEE Sensors Journal, vol. 22, no. 19, pp. 18218-18227, 1 Oct.1, 2022, doi: 10.1109 / JSEN.2022.3198395.

[0011] AG Stove and SR Sykes, "A Doppler-based target classifier using linear discriminants and principal components," 2003 Proceedings of the International Conference on Radar (IEEE Cat.No.03EX695), Adelaide, SA, Australia, 2003, pp. 171-176, doi: 10.1109 / RADAR.2003.1278734.

[0012] X. Feng, X. Hu and Y. Liu, "Radar signal sorting algorithm of k-means clustering based on data field," 2017 3rd IEEE International Conference on Computer and Communications (ICCC), Chengdu, China, 2017, pp. 2262-2266, doi: 10.1109 / CompComm.2017.8322938.

[0013] J. Lee, SK Yoo,"Radar-Based Detection of Respiration Rate with Adaptive Harmonie Quefrency Selection," Sensors, 2020, vol. 20, no. 6, p.1607. https: / / doi.org / 10.3390 / s20061607.

[0014] Chioukh, L. (2009). Integrated medical radar system for precision monitoring of heartbeat and respiratory status [Master's thesis, École Polytechnique de Montréal].

Claims

A method for detecting the posture of an individual comprising the steps of: [100] continuously emitting coherent signals, [200] continuously collecting the echoes of the signals emitted in step [100] and extracting the corresponding beat signals, [300] extracting from said beat signals the distance-Doppler signals, [400] assigning to each potential target a 3D position, [500] determining whether the potential target corresponds to the presence of an individual and assigning it a state if so, characterized in that step [500] determining whether the potential target corresponds to the presence of an individual and assigning it a state if so, comprises a step [520] consisting of using a statistical test to compare the distribution of the signals corresponding to the potential targets to a reference distribution. Method according to the preceding claim, characterized in that step [500] determining whether the potential target corresponds to the presence of an individual and assigning it a state if necessary, further comprises the step: [530] grouping a plurality of 3D positions corresponding to an individual and analyzing the geometric and statistical distribution of the plurality of 3D positions and assigning a posture to the latter. Method according to the preceding claim, characterized in that step [530] grouping a plurality of 3D positions corresponding to an individual and analyzing the geometric and statistical distribution of the plurality of 3D positions and assigning a posture to the latter, further comprises an analysis of the Doppler signals associated with said plurality of 3D positions. Detection method according to one of the preceding claims, characterized in that said posture assigned in step [530] grouping a plurality of 3D positions corresponding to an individual and analyzing the geometric and statistical distribution of the plurality of 3D positions and assigning a posture to the latter, is chosen from the group consisting of “static standing”, “standing while walking”, “sitting” and “lying down”. Method according to one of the preceding claims, characterized in that it further comprises the step: [540] comparing the successive postures attributed in step [530] grouping a plurality of 3D positions corresponding to an individual and analyzing the geometric and statistical distribution of the plurality of 3D positions and assigning a posture to the latter, and assigning a state to said individual. Detection method according to the preceding claim, characterized in that said step [540] comparing the successive postures attributed to step [530], is carried out by measuring the Mahalanobis distance. Detection method according to one of the preceding claims, characterized in that said state attributed to step [540] comparing the successive postures attributed to step [530], is chosen from the group consisting of “static”, “movement”, “fall” and “no fall”. Method according to one of the preceding claims, characterized in that the statistical test used in step [520] consisting of using a statistical test to compare the distribution of the signals corresponding to the potential targets to a reference distribution, is a Kolmogorov-Smirnov test. Method according to one of the preceding claims, characterized in that step [300] extracting the distance-Doppler signals from said beat signals is carried out by distance processing [310] then by Doppler processing [320]. Method according to one of the preceding claims, characterized in that step [400] assigning a 3D position to each potential target, comprises: [410] a step of extracting the distances of said potential targets, and [420] a step of extracting the elevation and azimuth angles of said potential targets. Method according to one of the preceding claims, characterized in that it further comprises a step: [600] measuring the respiratory and cardiac parameters of said individual. Method according to the preceding claim, characterized in that step [600] measuring the respiratory and cardiac parameters of said individual, is carried out from the Doppler signals associated with said individual. Device for implementing a method according to one of the preceding claims, characterized in that it comprises a coherent radar and a computer programmed so as to carry out the steps of the method according to one of the preceding claims. Device according to the preceding claim, characterized in that said radar is fixed at a height of between 150 and 250 cm in an adjustable manner. Device according to one of claims 13 or 14, characterized in that said radar operates at a wavelength suitable for the detection of human tissue and with a bandwidth allowing a spatial resolution of less than 1 meter.

Images