WO2018115255A1

WO2018115255A1 - Method for autonomously geolocating a person traveling on foot or by means of a non-motorized vehicle and associated device

Info

Publication number: WO2018115255A1
Application number: PCT/EP2017/084040
Authority: WO
Inventors: Gentian Jakllari; André-Luc BEYLOT
Original assignee: Institut National Polytechnique De Toulouse; Centre National De La Recherche Scientifique
Priority date: 2016-12-21
Filing date: 2017-12-21
Publication date: 2018-06-28
Also published as: EP3559598A1; FR3060760B1; FR3060760A1

Abstract

The invention relates to a method for autonomous geolocation and an associated device for determining the position of a person traveling on foot or by means of a non-motorized vehicle using inertial data measured while the person is traveling. An elementary mode of travel is recognized by comparing the inertial data measured with predetermined signatures, each signature being previously associated in a unique manner with an elementary mode of travel independently of the person in question. The speed and direction of the travel of the person are then estimated according the recognized elementary mode of travel and according to all or part of the measured inertial data. The position of the person is then evaluated according to the estimated speed and direction of travel and the position evaluated during a previous iteration.

Description

Autonomous geolocation method for a person traveling on foot or by means of a non-motorized machine and associated device

1. Field of the invention

The present invention is in the field of geolocation. It relates more particularly to a method and an associated device for autonomously locating a person moving on foot or by means of a non-motorized traveling device. It finds a particular application for the geo-location of pedestrians moving inside buildings.

2. Prior Art

There are currently multiple geo-location systems relying on communications infrastructure, such as wireless (Wi-Fi, Li- Fi) and mobile wireless (3G, 4G, 5G) and / or Global Positioning System (GPS) satellite networks for determining the position of a mobile terminal on the basis of received signals.

However, such infrastructures are not suitable for electro-magnetically confined or disturbed environments; this is the case inside buildings (eg buildings, ships) or in highly urbanized outlying areas (eg city center) where the reception of GPS signals can be very poor quality, or even impossible due to attenuations, any kind of interference or electromagnetic interference. For example, the reception of GPS signals can be made impossible outdoors in highly urbanized areas, particularly because of the phenomenon of urban canyoning due to destructive interference in the presence of high facades.

To respond to this problem, it may be envisaged, under certain conditions, to install in such environments, points or wireless access points, for example of the Wi-Fi type, for geolocating a mobile terminal. Thus, the position of a mobile terminal can be determined either by the terminal itself by exploiting the signals received from one or more access points or by the access point system from the signals received from the terminal. mobile terminal. In all cases, cooperation is required between the mobile terminal and the infrastructure including the access points. This type of solution is particularly disadvantageous because it requires the installation and / or maintenance of communicating infrastructures, which can be very costly and restrictive, especially since the areas to be covered by the access points can to be geographically very extensive. Moreover, in certain areas, it is not desirable or authorized to deploy new sources of radio frequency emissions, in particular to limit electromagnetic pollution.

This type of solution also requires a learning phase called fingerprinting, specific to each environment, an expensive task that must be repeated each time the infrastructure is updated.

Therefore, there is currently a real need to geo locate, for example by means of a simple mobile terminal, a pedestrian completely independently, that is to say without the need to use any external infrastructure (eg satellites, wireless access points, radio-mobile base stations).

Autonomous stand-alone geolocalization techniques called dead-reckoning have been proposed. They consist in calculating, step by step, the positions of the pedestrian in displacement, starting from a known reference position, on the basis of measurements of inertial data carried out in real time. However, these techniques suffer from an integration drift phenomenon inherent in the technology of inertial sensor data currently available on the market, having the effect of rendering a wrong result after a certain distance traveled. . Currently, the implementation of these techniques on a mobile phone does not provide a reliable result, beyond a dozen meters traveled.

As a result, periodic recalibration must be performed, for example by building on existing Wi-Fi infrastructure, a combination of crowdsourcing elements and virtual landmarks, and prior knowledge of geometry. specifies places.

The need to perform a re-calibration every 12 meters is not only restrictive but also resource consuming and makes the geolocation technique dependent on external elements, which does not make the geolocation technique completely autonomous.

Autonomous techniques have been proposed based essentially on counting the number of steps. However, in this case, a calibration operation having to estimate the average size of a stride of the pedestrian, ie the average distance traveled between two steps is necessary. This calibration is required before each use. In addition, this calibration is specific to each pedestrian considered.

These techniques are not only restrictive, insofar as they require an initial calibration specific to each user, but also have a limited geo-localization accuracy that can not be attained in the types of environments targeted, especially inside. of buildings, geolocalisation details sufficient to allow to guide efficiently and without error a pedestrian during the trip.

Indeed, for a given user in a given outfit and environment, the size of a stride is likely to vary from one step to another. This variability is a source of inaccuracy in the estimation of the distance traveled. This inaccuracy will be all the more important as the number of steps taken by the pedestrian during his journey is high, so that a periodic calibration is necessary. For example, using sensors embedded on a mobile phone, a recalibration is necessary, on average every 12 meters traveled.

The same applies when a person moves using a non-motorized transport device, such as a bicycle, a tricycle, a scooter, skates / skis / skateboard or a gliding device, such as only ice skates, skis.

Thus, there is a need to provide a fully autonomous, reliable and highly accurate location-based geolocation solution that is easy to implement and does not require any initial calibration before use, nor recalibration during use.

3. Object of the invention

The present invention aims to remedy the aforementioned drawbacks by proposing a technical solution for geolocating in a simple, autonomous and precise manner, a person walking on foot (ie pedestrian) or by means of a non-motorized machine during his movements, especially in buildings.

These goals are achieved by an autonomous geolocation process of a first person traveling on foot or by means of a non-traveling craft. motor, said method intended to be implemented on a portable device integral with said person, said method comprising the following steps:

• measurement of inertial data relating to a displacement of said first person;

Recognizing a basic mode of movement of said first person, in which all or part of the measured inertial data are compared to predetermined signatures, each signature being previously uniquely associated with a basic mode of movement independently of said first person considered;

Estimating the speed and direction of movement of said first person, according to the recognized elementary displacement mode and all or part of the measured inertial data; and

• evaluating the position of said first person, according to the estimated speed and direction of movement and the position evaluated during a previous iteration of the evaluation step.

The use of unique signatures advantageously avoids any prior step of calibration of the geo-location algorithm according to the first person, since each signature is universally valid regardless of the person considered.

Estimating the speed and position of the first pedestrian according to the elementary movement recognized in real time advantageously makes it possible to improve the accuracy of the evaluation of the position of a person traveling.

The method according to the invention is implemented on a device, portable or portable, integral with the person. By "solidary of the person" is meant that the device is held in the hand of the person, attached to the body thereof by any means hooked (eg type of mobile phone holder attached to the arm, etc.) or placed / kept in / on a garment worn by said person {eg in a trouser pocket, etc.).

According to one characteristic of the invention, the method comprises a first preliminary step, in which each signature is elaborated on the basis of at least one temporal signal of inertial data obtained for a given elementary mode of displacement and performed by at least a second person.

Since the development of signatures has been done beforehand for any type of displacement (eg walking, running), the geolocation process can be implemented for any person, without any calibration step, taking into account the uniqueness of the pattern associated with each elementary movement previously indicated.

Advantageously, the second person may be different from the first person, the signature is uniquely associated with a basic mode of movement, regardless of the person considered.

According to another characteristic of the invention, during the recognition step, the measured inertial data are compared to the signature elements of each of the predefined elementary displacements, by calculating a correlation score weighted by weights previously allocated to each type of inertial data, so that the recognized mode of travel is the one for which the correlation score is the highest.

According to another characteristic of the invention, during the recognition step, the inertial data are selected from data of vertical acceleration, frontal, transverse, rotation.

According to another characteristic of the invention, the vertical acceleration data are assigned a multiplicative factor constituting a preponderant weight compared to the other inertial data compared during the recognition step.

The inventors have found that among all the types of inertial data collected, the vertical acceleration data are particularly singular for the identification of a mode of displacement. The vertical acceleration data is particularly reliable for identifying a distinctive pattern associated with walking activity or running activity, regardless of the person considered.

According to another characteristic of the invention, the vertical acceleration data are assigned a first weight and:

· The frontal acceleration data is assigned a second weight of value at least half that of the first weight, and / or • the transverse acceleration data are assigned a third weight of value at least eight times lower than that of the first weight, and / or

• the roll type rotation data is assigned a fourth weight value at least half of that of the first weight, and / or · the pitch type rotation data is assigned a fifth value weight less than four times less than that of the first weight. The choice of these coefficients attributed to the respective inertial data is advantageous insofar as it makes it possible to obtain a recognition rate of a relatively high elementary displacement.

According to another characteristic of the invention, the inertial data compared during the recognition step correspond to a sample of temporal data with a duration of between 1 s and 2 s.

The acquisition of inertial data for one second is sufficient for the analysis since several patterns can be captured for a duration between 1s and 2s. In particular, a sample of duration equal to 1 s presents an excellent compromise between the speed of processing and the accuracy.

The fact that the signature signal extends over a duration greater than 1 s is not incompatible with a sample duration equal to 1 s since the search for the signatures in the measured signal is carried out in the frequency domain .

According to another characteristic of the invention, the speed of movement of the first person is estimated, during the estimation step, according to the recognized elementary displacement mode and a displacement rate of said first person estimated in real time from measured inertial data.

Taking into account the rate of movement of the person advantageously improves the accuracy of the estimated speed.

According to another characteristic of the invention, the evaluation of the speed of displacement is carried out on the basis of a model previously established and defining for each mode of elementary displacement a univocal relationship between the rate and the speed of displacement.

Thus, the model makes it possible to associate a range of cadence values with a range of displacement speed values, for each mode of displacement. elementary. The fact of previously identifying a subdomain of values of possible travel speeds for each elementary displacement mode makes it possible to reduce the range of possible values for the estimation of the actual speed, which is particularly advantageous for a processing operation. real time.

According to another characteristic of the invention, the model is produced by polynomial regression, on the basis of a set of discrete measurement points, each point associating a speed with a rate of displacement.

The regression allows to obtain a continuous relation between the speed of displacement and the rate from a set of discrete values for each mode of elementary displacement.

According to another characteristic of the invention, the estimation of the direction of movement of said first person comprises:

A filter sub-step in which all or part of the measured inertial data is filtered by subtracting corresponding data included in the signature of the recognized mode of displacement; and

A substep of correcting a systematic error inherent in the means for measuring said inertial data.

The combination of filtering and systematic error correction significantly improves the accuracy with which the person's position is determined.

According to another characteristic of the invention, the method comprises a step of measuring an altitude change such that a vertical displacement mode is identified if an altitude change greater than a threshold for a duration greater than a threshold predetermined is detected, said vertical displacement mode being determined at least according to an elevation parameter, Le. of change of altitude with respect to a predetermined threshold.

Advantageously, the predetermined threshold is set so as to correspond to a significant change in altitude for detecting a change of stage during the displacement. According to another characteristic of the invention, a horizontal component of the displacement is estimated as a function of a vertical component of the displacement and a reference angle, preferably equal to 30 °.

This feature is particularly advantageous for estimating geolocation points, for example, in the case where the user climbs stairs, in particular so-called escalators or spiral staircases.

The invention also relates to an autonomous geolocation device for a person moving on foot or by means of a non-motorized moving device, said device comprising:

Measuring means for measuring inertial data relating to a displacement of said person;

Recognition means for recognizing a basic mode of movement of said person, said recognition means being adapted to compare all or part of the inertial data measured at predetermined signatures, each signature being previously stored in storage means in association with unique to a particular mode of movement regardless of the person considered;

Estimating means for estimating the speed and direction of movement of said person according to the recognized elementary displacement mode and the measured inertial data; and

Evaluation means for evaluating the position of said person according to the estimated speed and direction of movement and the position evaluated during a previous iteration of the evaluation step.

The invention also relates to a computer program comprising instructions adapted to the implementation of at least one of the steps of the method according to the invention as described above, when said program is executed on a computer, in in particular, a mobile terminal such as a mobile phone, a tablet or any system comprising an inertial measurement unit (or IMU: Inertial Measurement Unit). This program can use any programming language, and be in the form of source code, object code, or intermediate code between source code and object code, such as in a partially compiled form, or in any other form desirable shape.

The invention also provides a means for storing information, removable or not, partially or completely readable by a computer or a microprocessor comprising code instructions of a computer program for the execution of at least one of steps of the processes according to the invention as described above.

The information carrier may be any entity or device capable of storing the program. For example, the medium may comprise storage means, such as a ROM (Read Only Memory), for example a microcircuit ROM, or a magnetic recording means, for example a hard disk, or a memory flash.

On the other hand, the information medium may be a transmissible medium, such as an electrical or optical signal, which may be conveyed via an electrical or optical cable, by radio or by other means. The program according to the invention may in particular be downloaded to a storage platform of an Internet type network.

Alternatively, the information medium may be an integrated circuit, in which the program is incorporated, the circuit being adapted to execute or to be used in the execution of the method in question.

The aforementioned information carrier and computer program have characteristics and advantages similar to the method they implement.

One of the ideas underlying the invention is to evaluate the position of a person moving on foot or by means of a non-motorized moving device, according to a basic mode of movement identified by a single signature previously determined independently of the person concerned.

For example, in the case of a person walking on foot (The pedestrian), we can distinguish the following modes of travel: normal walking, walking health, walking in a hurry, running.

The fact of being able to identify each mode of elementary displacement by a single signature whatever the person considered is a characteristic in-depth analyzes conducted by the inventors. More specifically, they have succeeded in highlighting that the inertial data measured during the displacement of the person, for example by means of conventional sensors embedded in a mobile phone, a tablet or any other device comprising an inertial unit, present particular temporal patterns specific to each mode of elementary displacement, whatever the person considered. In other words, each of these patterns constitutes an invariant, regardless of the time signal of the inertial data measured for the same elementary mode of displacement.

Thanks to its unique signature, each elementary mode of displacement can be recognized by comparing in real time the acquired inertial data with respect to a set of uniquely predetermined reference signatures for each elementary mode of displacement.

The inventors have found that, among the set of measured inertial data, the vertical acceleration is that which presents the most significant characteristic patterns, in particular to distinguish more finely and reliably the different activities between them and in particular of walking and race.

For each elementary mode of displacement, the inventors have also demonstrated that, according to the analysis of their experimental results, taking into account the pedestrian movement rate, ie its number of steps per second, in the determination. the speed of displacement advantageously makes it possible to improve the geolocation accuracy of the method. The inventors have developed a model making it possible to determine, for each mode of elementary displacement, the speed of displacement as a function of said rate. 4. Brief description of drawings

Other features and advantages of the invention will become apparent in the description below, in relation to the accompanying drawings, given by way of non-limiting examples, and in which:

Figure 1 schematically illustrates a particular embodiment of the device according to the invention;

Figure 2 illustrates a particular embodiment of the method according to the invention; Figure 3 illustrates an example of inertial data signals measured in the method of the invention;

Figure 4 illustrates a model connecting the speed of movement to the rate of movement of a pedestrian according to a feature of the invention; and - Figure 5 schematically illustrates another particular embodiment of the invention for geo-localization in three dimensions.

The invention is described below with reference to Figures 1 to 5 in the context of the geolocation of a pedestrian on the move.

Figure 1 schematically illustrates the hardware architecture of a device 100 according to a particular embodiment of the invention. This device is adapted to implement the steps of the method according to a particular embodiment of the invention.

This device is intended to be integral with the pedestrian throughout his travels. It can be held in a pedestrian's hand or attached to a part of its body by any means of stable fixation.

The device 100 comprises measuring means 1 adapted to measure in real time inertial data, such as angular velocity data and / or linear and / or angular acceleration. In the present example, these measuring means 1 consist of an integrated inertial data sensor 1.

The integrated inertial data sensor 1 comprises at least one gyrometer 10.

In the present example, it is a triaxial gyrometer adapted to measure an instantaneous angular velocity represented by a vector ω having three components of rotational speed co _x , co _y , co _z around each of the axes of the gyrometer . Thus, the gyrometer 10 is adapted to measure the yaw or heading speed (yaw), the pitching speed (pitch) and the rolling speed (mil), each speed being expressed in Rad / s.

The integrated inertial data sensor 1 comprises an accelerometer 12. In the present example, it is a tri-axis accelerometer adapted to measure a local linear acceleration represented by a vector γ having three linear acceleration components γ _χ , y _y , γ _ζ along each axis of Γ accelerometer respectively. Thus, Γ accelerometer 12 is able to measure the vertical acceleration, the frontal acceleration and the transverse acceleration, each acceleration being expressed in m / s ² . More generally, it is considered that there are in total k sensor elements, where k denotes a natural integer, these elements may belong to the same sensor or to different sensors depending on the embodiment considered. In the present example, there are six sensor elements: k = 6 (the three for the accelerometer 12 and three for the gyrometer 10).

Subsequently, the inertial data will generally denote acceleration or velocity. More particularly, the terms "inertial magnitude" or "type of inertial data" will be used interchangeably to designate a linear acceleration (Le. In translation) along one of the three axes {x, y, z} of an orthonormal coordinate system of reference or an angular velocity (rotation) about one of the three axes {x, y, z} of a reference orthonormal reference.

By way of illustrative example, in the particular case where k = 6, there are 6 types of inertial data or 6 inertial variables selected from: a linear acceleration along the x axis, a linear acceleration along the y axis, a linear acceleration along the z axis, an angular velocity along the x axis, an angular velocity along the y axis, an angular velocity along the z axis.

For the sake of simplification, only a gyrometer 10 and an accelerometer 12 have subsequently been considered for illustrative purposes. However, one skilled in the art will of course be able to adapt the present invention to the case where several accelerometers and / or several gyrometers are jointly embedded within the device 100.

The device 100 further comprises:

a central processing unit 4 comprising a microprocessor; a random access memory 6 of the Random Access Memory (RAM) type and a Read Only Memory (ROM) type 5;

an input / output port 7 (I / O: Input / Output);

a communication interface 8, for example of the short-range radio frequency type such as a Bluetooth® interface; and

optionally, a pressure sensor 9.

All the constituent elements of said device 100, as described above, can be typically integrated on the same printed circuit board PCB type monobloc. The read-only memory 5 constitutes a recording medium within the meaning of the invention. This medium 5 stores a computer program PG1 able to implement, when executed by the central unit 4, the geolocation process steps performed by the device 100 according to the invention, as illustrated in FIG. 2.

Subsequently, it is considered as an example, that the device 100 is constituted by a mobile phone type smart phone currently available on the market and in which are embedded a tri-axis gyrometer 10 and a tri-axis accelerometer 12 sub the shape of the integrated inertial data sensor 1.

However, the device according to the invention is obviously not limited to a mobile phone but is aimed at any type of device capable of autonomously measuring inertial variables such as acceleration and / or speed of rotation. For example, it may be an integrated sensor comprising an inertial or inertial unit (IMU).

According to an alternative embodiment, the measuring means 1 (ie inertial sensors) are embedded on the device, while the microprocessor of the processing unit 4 is dissociated from the device 100. This variant embodiment is particularly suitable in the case where the treatment is postponed or displaced. In this case, all or part of the post-treatments can be delegated to one or more fixed or mobile devices themselves. For example, this variant could be advantageously used to determine the round of a guard in a supermarket after he has completed his round.

FIG. 2 illustrates a particular embodiment of the method according to the invention, as implemented by the device 100 illustrated in FIG.

The present invention will now be described in a particular embodiment, in which we consider a pedestrian who moves by walking or running according to different modes of elementary movement.

During a first preliminary step E01, one of the signals Ai, _k (t) constituting inertial data is acquired by means of one or more devices according to the invention 100. These inertial data are represented in the form of signals such as those represented by way of example in FIG. 3. These inertial data signals are measured in the time domain, for each mode of measurement. elementary displacement i, from sensor elements k, where i and k denote natural numbers such that l <i <m and l <k <p, where m denotes the total number of elementary modes of movement and p denotes the number total of sensor elements considered.

The measurement of these signals is carried out for one or more pedestrians (second person) any that do not necessarily correspond to the pedestrian end user of the invention (first person), that is to say that one wishes to geolocate.

In the present example, we consider that m = 4 to designate the following four elementary modes of movement: walking, strolling, rushing, jogging. It is considered that p = 6 to designate the three elements of the accelerometer 12 measuring the three acceleration components and the three elements of the gyrometer 10 measuring the three components of angular velocity (roll, pitch, yaw).

During the first prior step E01, each measured signal A, _k (t) is truncated, centered and completed with stuffing bits "0", before being converted into the frequency domain. This conversion is effected, for example, by a Fast Fourier Transform (FFT) applied to the signals A _iik (t), so as to obtain a set of signatures in the frequency domain: Si, _k = FFT ( A, _k ) where Si, _k designates a signature element obtained by FFT on the signal measured by the sensor element k for the elementary displacement mode i.

All the signals S ^ _k are normalized and stored in a signature matrix X = [Si, _k ] where l <i <m and l <k <p. In the present example, the matrix X includes for each of the four elementary displacement modes considered (i = 1: walk, i = 2: walk of health, i = 3: walk in a hurry, i = 4: run), a signature consisting of six signature elements as listed in the table below.

Mode Element No. 1 Mode No. 2 Mode No. 3 Mode No. 4 Sensor (k) (i = 1) (i = 2) (i = 3) (i = 4)

k = 1 SU S21 S31 S41 k = 2 S12 S22 S32 S42 k = 3 S13 S23 S33 S43 k = 4 S14 S24 S34 S44 k = 5 S15 S25 S35 S45 k = 6 S16 S26 S36 S46

Table: Signature Matrix

Thus, the signature S1 comprises the six signature elements SU, S1, S13, S14, S15, S16 making it possible to uniquely characterize a first mode of elementary displacement (i = 1), independently of the physical parameters specific to the considered pedestrian such as height, weight or sex. The same goes for each of the other modes of elementary displacement.

It will be noted that the universal nature of these signatures is particularly advantageous compared to solutions of the prior art that require calibration as a function of one or more physical parameters of the pedestrian.

Once the signature matrix is generated, it is stored in the read-only memory 5 of the mobile phone 100 and can be directly used to evaluate the position of a pedestrian according to the steps of the method of the invention, as described. hereinafter, regardless of the pedestrian considered and without requiring any calibration operation.

In the present embodiment, the first preliminary step E01 is implemented by the processor 4 of the mobile phone 100. However, in other embodiments, the processes performed on the signals Ai, _k (t) for the obtaining the X signature matrix can be performed centrally by at least one processor on at least a remote server (not shown), to which the mobile phone of the pedestrian can connect via the Internet and / or a mobile network, to download the matrix X. According to one embodiment, the X signature matrix may have been pre-recorded in the memory of the pedestrian's mobile phone. In this case, the X signature matrix can be used directly by the phone without requiring prior connection. By default, it is assumed that the pedestrian is inactive, so that the elementary mode of displacement is initially fixed as being of type "Idle".

During a measurement step El, the gyrometer 12 measures the angular velocities of roll, pitch, yaw, while the accelerometer 10 measures the vertical, frontal, transverse accelerations. These measurements are stored temporarily in the RAM 6 of the mobile phone 100.

In general, the method according to the invention is executed automatically when the conventional means of geolocation of the mobile terminal on which is implemented the process are no longer able to acquire geolocation data.

This is particularly the case when the mobile terminal enters a building, within which signals necessary for geolocation, such as GPS signals, are no longer detected by the terminal. In this case, the instantaneous position being not determinable by the conventional means of geolocation, the measurements of the inertial data are performed during the measurement step El. As an illustrative example, Figure 3 shows inertial data signals measured for walking (Fig. 3a) and running (Fig. 3b). These measurements were made in the case where the pedestrian holds his phone in the right hand, so that the phone is in solidarity with the pedestrian. For the sake of simplicity, only the yaw rotation speed expressed in Rad / s and the frontal and vertical accelerations expressed in m / s ² have been represented. In general, all or part of the inertial data provided by one or more sensor elements k can be considered.

According to one particularity of the invention, the vertical acceleration is considered with a preponderant weight compared to the other measured inertial variables to uniquely identify a mode of elementary displacement. This peculiarity results from the observations made by the inventors on the temporal evolution of the set of measured inertial data signals.

For a "normal" walking activity, the inventors have demonstrated, in the time domain, the repetition of a vertical acceleration motif M1 specific to this activity, this pattern being characterized by the presence of two acceleration peaks. vertical. As illustrated in FIG. 3a, these two peaks have a substantially equal amplitude and are separated by a time interval of between 50 and 150 milliseconds. The inventors have found that this characteristic pattern of walking appears between the instant t2 where the heel of one foot of the pedestrian strikes the ground and the instant t '1 where the toes of the other foot cease to be in contact with the ground. According to the inventors' analysis, this pattern of substantially invariant shape appears systematically for a walking activity, whatever the pedestrian considered. Therefore, this pattern M1 is considered a reliable indicator for recognizing a walking movement. For any race activity, the inventors have demonstrated in the time domain, the repetition of a vertical acceleration pattern M2 specific to this activity characterized by a plateau of saturation of the vertical acceleration of a duration between 50 and 150 milliseconds, the plateau being immediately preceded by a peak acceleration amplitude less than the plateau saturation level as shown in Figure 3b. The inventors have found that this invariant pattern appears between the instant t4 where the heel of a pedestrian foot hits the ground and the instant t'3 where the toes of the other foot cease to be in contact with the ground. This pattern substantially invariant form appears systematically for a race activity, regardless of the pedestrian considered. Therefore, this pattern M2 is considered a reliable indicator for recognizing a race movement.

In a general manner, during a recognition step E3, the processor implements the following substeps:

(i) comparing the measured inertial data with respect to the predetermined unique signatures (X), so as to obtain a set of coefficients (r (i, k)) for each of the predefined elementary displacements;

(ii) multiplying the coefficients by a respective multiplicative factor corresponding to a weight (W) previously assigned to an inertial quantity, so as to obtain a set of weighted values,

(iii) calculating a correlation score corresponding to the sum of the weighted values;

the recognized mode of travel being the one for which the correlation score is the highest.

During the recognition step E3, the processor 4 processes each inertial data signal measured in real time according to the following operations:

extraction of a sample of the acquired inertial data of a duration greater than 1 second (s), preferably between 1 s and 2 s, conversion of this sample in the frequency domain by application of a fast Fourier transform (FFT ) on the sample,

truncation of the converted sample so as to eliminate edge effects, bit stuffing to ensure that the size of the signal obtained is equal to a power of 2,

comparison in the frequency domain of the sample converted to each of the signatures contained in the X signature matrix.

The inventors have found that the acquisition of inertial data for a duration of between 1 s and 2 s, and in particular equal to 1 s, is sufficient to allow a reliable recognition of the elementary mode of displacement, since several characteristic patterns of this element mode of movement can be captured during this time.

In particular, the processing of a sample of a duration equal to 1 s presents an excellent compromise between the speed of processing and the accuracy at which the instantaneous position of the pedestrian is evaluated. Indeed, the inventors have determined, on the basis of measurements made over a path of 200 m, that the position of the pedestrian can be measured with an accuracy of at least 90% in the case where the duration of the sample on which carries the recognition step is equal to 1 s. When the sample duration is extended to 2.5 s, this accuracy approaches 100%.

The fact that the signal used to obtain the signature extends over a duration greater than 1 s is not incompatible with a sample duration equal to 1 s since the search for the signatures in the signal measured performs in the frequency domain.

As an illustrative example, the comparison of the sample of the measured signal with respect to the unique signatures is performed by the processor 4. For this, each sample Z is compared to the set of signatures of the matrix X by calculating the coefficient of Pearson r _pea rson according to the following formula:

_ X. Z ^T - nX Z

Ipearson ^- δ where # () = _ [Σ _{= 1} X] - nX ^{2 where} n denotes the size of the vector X

^T designating the matrix transpose and denoting the average. Thus, we obtain a matrix of Pearson coefficients r, whose elements are defined by r (i, k) = r _pea rson (X (i, k), Z (k)) for each elementary displacement mode i. In the present example, each sample Z is compared to a set of six signature elements for each of the four elementary modes of displacement, so as to obtain a set of 6 × 4 = 24 Pearson coefficients.

A scorejoc scalar correlation score is then calculated by the processor 4 by multiplying each of the Pearson coefficients by a multiplicative factor corresponding to the respective weight previously associated with each inertial magnitude.

According to one particularity of the invention, a weight predominant is assigned to the vertical acceleration data in view of the fact that the inventors have found experimentally that this magnitude is a reliable indicator for distinguishing a mode of elementary displacement.

According to another feature of the invention, the frontal acceleration data is assigned a weight two times lower than the weight assigned to the vertical acceleration data, the transverse acceleration data are weighted eight times less than the weight assigned to the vertical acceleration data, the roll-type rotation data is weighted half as much as the weight assigned to the vertical acceleration data, the pitch-type rotation data is assigned weighing four times less than the weight assigned to vertical acceleration data.

In this example, the multiplicative factors 0.4 are assigned; 0.2; 0.05; 0.2; 0.1; 0.05 respectively to the data of vertical acceleration, frontal, lateral and speed of rotation of yaw, pitch, roll. These multiplicative factors, or weights, are stored in the form of a matrix of weight W = {0.4; 0.2; 0.05; 0.2; 0.1; 0.05}.

This choice of these values is particularly advantageous insofar as it makes it possible to reliably determine an elementary mode of displacement (i.e. relatively high recognition rate).

Given the low weight assigned to the rotational speed of roll (mil), i.e. 0.05, we can omit to consider this inertial magnitude in the calculation of the score in order to simplify the calculation. Thus, according to an alternative embodiment, the Pearson coefficient r (i, 6) is not calculated for this quantity (k = 6) and in this case, the weight matrix is reduced to W = {0.4; 0.2; 0.05; 0.2; 0.1} where p = 5.

In general, for each sample, the scalar correlation score score_loc is obtained for each mode of displacement i by calculating the sum weighted Pearson coefficients as follows:

where ^T designates the matrix transpose.

In the present example, we have for each sample a set of 6 Pearson coefficients stored in the form r (i) = (r (i, l), r (i, 2), r (i, 3) , r (i, 4), r (i, 5), r (i, 6)} for a mode of displacement 1. To account for the weights assigned to the different measured inertial quantities, we multiply by:

• 0.4 the Pearson coefficient relative to the vertical acceleration data r (i, l),

• 0.2 the Pearson coefficient for the frontal acceleration data r (i, 2),

• 0.05 the Pearson coefficient relative to the lateral acceleration data r (i, 3),

• 0.2 the Pearson coefficient for yaw rate data r (i, 4),

• 0.1 the Pearson coefficient relative to the pitch rotational speed data (i, 5),

• 0.05 the Pearson coefficient relative to roll speed data r (i, 6).

The sum of these weighted values is calculated as follows:

score_loc = 0.4.r (i, l) + 0.2.r (i, 2) + 0.05.r (i, 3) + 0.2.r (i, 4) + 0, 1. r (i, 5) + 0,05.r (i, 6).

If the score obtained for the displacement mode i is sufficiently high, for example, if its value is greater than 0.4 (i.e.loc score> 0.4), it is considered that this mode of displacement i can be potentially retained. In this case, we save this score as a reference score (score = s core_loc) to which will be compared the calculated score during a next iteration of the calculation algorithm for another elementary mode of displacement.

For example, during a first iteration, the score of the sample Z is calculated with respect to the walking activity (i = 1):, score_ / Oc = r (1) ^T xW where r ( l) (k) = r _pearson (X (l, k), Z (k)). If this score is greater than 0.4 (score_loc> 0.4), then we assign the value of scorejtoc to a score variable. In a second iteration, the score of the sample Z is calculated with respect to the walking health activity (i = 2): score, c = r (2) ^T xW or r (2, k) = r _pearson (X (2, k), Z (k)). If this score is greater than 0.4 then we assign the scorejtoc value to the score variable. This is done in succession for each predefined elementary displacement mode, so that the elementary displacement which has reached the highest score for the activity analyzed will be considered as finally recognized. If no iteration makes it possible to obtain a score greater than 0.4, then no elementary mode of movement is recognized. In this case, it is considered that the pedestrian is not in motion (idle).

During an estimation step E5, the speed and direction of movement of the pedestrian are estimated according to the recognized elementary displacement mode and the inertial data measured in real time by the gyrometer 10 and / or the accelerometer 12.

According to one particularity of the invention, the estimation of the direction of displacement comprises a substep of filtering E52 of all or part of the measured inertial data. This filtering substep E52 aims to remove from the measured signal spurious data corresponding to body movements, such as arm swing or lateral waddling. The inventors have found that such movements can constitute a source of inaccuracy for the determination of the direction of displacement and that the recognition of a mode of elementary displacement can be advantageously used to extract these parasitic data.

In this example, filtering is considered to apply to the measured yaw velocity values. These values are corrected during the filtering sub-step E52 by the processor 4 as a function of the signature of the recognized elementary displacement mode. For this, the yaw rate values measured by the gyrometer 10 are converted by FFT into the frequency domain, from which the corresponding values of the signature (i.e. signature elements) of the recognized elementary displacement mode are subtracted. More specifically, the spectral component of the converted measurement signal having a maximum amplitude is subtracted from the spectral component of maximum amplitude of said signature. Thus, the filtered version of the yaw rate resulting from filter substep E52 is an improved indicator of actual directional changes.

According to one particularity of the invention, the estimation of the direction of displacement comprises a substep of correction E54 of a systematic error (bias) and of noise inherent to the measuring means used (eg gyrometer 10). In the present example, the yaw rate 6> _m (t _n ) measured and filtered at time t _n as described above is expressed as follows:

e _m (t _n ) = è (t _n ) + b (t _n ) + μ {ί _η )

where Ô (t _n ) denotes the correct value of yaw rate, b (t _n ) denotes the systematic error and μ (ί _η ) denotes a Gaussian white noise at time t _n . The systematic error is estimated by processor 4 as follows:

n-l

b (t _n ) = nx E [è _m (ti) - é Ct _f -J] - ^ b (ti)

i = l

where E denotes the mathematical expectation.

By combining the E52 filtering and E54 correction sub-steps as described above, the inventors have succeeded in demonstrating, on the basis of experimental data, that the geolocation accuracy is significantly improved, compared to the case where only the sub-step of filtering is applied or in case none of the filtering and correction sub-steps is applied.

According to one particularity of the invention, the speed of movement of the pedestrian is estimated, according to the recognized elementary movement mode and a pedestrian movement rate estimated in real time from the measured inertial data.

The pace of movement of the pedestrian corresponds to the number of steps detected per second. The rate is estimated in real time and reliably, by converting the vertical acceleration measurements in the frequency domain, for example by applying an FFT-type transform, and by identifying the frequency of the spectral component having the amplitude. the highest.

According to one particularity of the invention, the speed of displacement is evaluated according to a model defining, for each mode of elementary displacement, a univocal relationship between the rate and the speed of displacement. The results of the experiments carried out by the inventors have made it possible to demonstrate that taking into account the rate in the evaluation of the speed of displacement advantageously makes it possible to appreciably improve the accuracy of geolocation.

Indeed, this model takes into account the fact that the speed of movement of a pedestrian can vary substantially within the same elementary mode of movement. This model is shown graphically as an illustrative example in Figure 4 for the following four elementary modes of motion: walking (strolling), walking (walking), hustling (running), running (jogging). As illustrated, for each elementary displacement mode, a continuous variation domain of the displacement velocity values is determined according to a subdomain of values of the estimated rate. For example, the speed of movement varies from 0.6 m / s to 1.3 m / s for cadence values between 1, 2 and 1, 7 steps / sec in the case of normal walking.

This model is obtained during a second preliminary step E02 performed before the implementation of the previously described calculation steps to geolocate the pedestrian. It should be noted that the development of this model is not relative to the pedestrian considered, Le. for which one wishes to determine the position.

The second preliminary step E02 consists, for each mode of elementary displacement, firstly to measure the speed and the rate using precise measurement methods and then to connect the measurement points in a continuous manner, by a known regression technique. implemented by a processor of a computer.

As an illustrative example, the rate can be measured reliably by several conventional mobile phones carried by volunteer pedestrians moving in the same elementary mode of travel (eg normal walking) over a reference distance (eg 20 meters) . The speed of movement is calculated as a function of the time taken by the pedestrian to travel this reference distance, this time being measured by means of a stopwatch.

As an illustrative example, the regression technique selected to continuously connect the measurement points is a polynomial regression technique. As illustrated in FIG. 4, the measurement points are identified, for each of the four elementary modes of displacement, by respective symbols, these points being connected by a curve obtained by polynomial regression.

It should be noted that, when this model is developed, it is stored in the read-only memory 5 of the telephone 100 so that it can be directly used to evaluate the position of a pedestrian according to the invention, regardless of the pedestrian considered. Once this model is established, it can advantageously be applied to any pedestrian considered to improve the accuracy of its geolocation, without requiring calibration pedestrian considered.

A centralized update of this model can be implemented on a remote server, to which the mobile phone intended to implement the invention can connect via the Internet and / or a mobile network, to download said model.

The inventors have demonstrated on the basis of experimental tests that the implementation of the geolocation method according to the invention in which the speed of displacement is estimated according to the rate according to the model described above allows to achieve a location accuracy of about 1 meter over distances of several hundred meters.

During an evaluation step E7, the position of said first person is evaluated according to the estimated speed and direction of movement and the position evaluated during a previous iteration of the evaluation step.

In particular, during the first iteration of the method according to the invention, the position "evaluated during a previous iteration of the evaluation step" must be understood as being a reference position, such as the position recently acquired by conventional geolocation means. Thus, when the method according to the invention begins to be executed, this reference position, which is not the subject of the invention, is known.

It will be noted that the method according to the invention relates to the evaluation of an instantaneous position independently, Le. no longer using conventional geolocation means, regardless of how the initial reference position is acquired. For example, this position can be acquired by any means of radio frequency reception, such as a GPS receiver included in the mobile terminal of the user or a communication interface of Wi-Fi, Bluetooth or Near Field Communication adapted to receive of an issuer the information of the reference position, for example, at a point of passage, such as the entrance of a building. The process steps described above make it possible to geolocate a pedestrian in a two-dimensional space forming a horizontal plane, i.e. parallel to the ground on which the building is supported.

Another particular embodiment of the invention is now described with reference to FIG. 5 to determine a vertical mode of displacement, which is particularly advantageous for continuing to geolocate in a reliable and autonomous manner the pedestrian when the latter changes from floor to floor. interior of the building.

According to another characteristic of the invention, a vertical displacement mode is recognized, as a function of at least one elevation parameter measured in real time.

For this purpose, the pressure sensor or barometer 9 of the device 100 measures at regular time intervals the pressure P. On the basis of these measurements, the processor 4 of the terminal 100 calculates an altitude differential Δζ such that Az = -AP / (pg), where ΔΡ denotes the measured pressure differential, p denotes the density of the air and g denotes the Earth's gravity. The processor 4 is adapted to detect an altitude change with respect to one or more predetermined thresholds (step E30).

For example, the processor 4 is adapted to detect a first altitude change greater than a first predetermined threshold, corresponding for example to an elevation of 1 m for a duration of 2 s.

As soon as an altitude change exceeding the first threshold is detected, a vertical displacement mode can be recognized in the following different cases.

If the processor 4 recognizes no movement of the pedestrian during the recognition step E3 but determines that the measured vertical acceleration a is greater than a predetermined acceleration threshold, for example equal to 1 m / s ² (ie 1 m / s ² ) then he deduces that the pedestrian moves vertically using an elevator (step E321). In this case, the coordinates of the user in a horizontal plane (ie parallel to the ground) remain unchanged during this vertical movement.

If the processor 4 recognizes no movement of the pedestrian during the recognition step E3 but detects that the measured vertical acceleration a is below said predetermined acceleration threshold (ie <1 m / s ² ), then the processor 4 determines that the pedestrian is standing on an escalator (E322). During this movement, the coordinates of the pedestrian vary not only vertically (that is to say perpendicular to the ground), but also in a horizontal plane.

Taking into account the angle at which the ramp of the escalator is inclined relative to the building floor, for example 30 °, the processor 4 can continue to accurately estimate the coordinates of the pedestrian in a horizontal plane.

If the processor 4 jointly detects an elementary movement of the pedestrian during the recognition step E3 and a rapid change of altitude Δζ (step E34), that is to say greater than a second predetermined threshold (eg Δζ> 0 , 5 m / s), then the processor 4 determines that the pedestrian moves on an escalator in operation (step E341) according to the elementary movement mode identified during the recognition step E3. Taking into account the angle at which the ramp of the escalator is inclined relative to the building floor, for example 30 °, the processor 4 can continue to accurately estimate the coordinates of the pedestrian in a horizontal plane.

If the processor 4 jointly detects an elementary displacement of the pedestrian during the recognition step E3 and a low altitude change Δζ (step E34), that is to say less than the second predetermined threshold (eg Δζ <0, 5 m / s), then the processor 4 determines that the pedestrian moves on a stair (step E342) according to the elementary movement mode identified. In this case, if the processor 4 further determines that changes of direction are fast and regular, then it deduces that the pedestrian moves along a spiral staircase. Taking into account the angle of inclination of the staircase (e.g. 30 °), the processor can continue to accurately determine the coordinates of the pedestrian in a horizontal plane.

In each of the cases described above, the method according to the invention makes it possible to precisely locate the movement of the pedestrian in a space in three dimensions, that is to say moving vertically.

The present invention is not limited to the case of the pedestrian. It also applies to anyone traveling by means of a non-motorized mobility device.

Non-motorized transport means, in particular, any machine propelled by the muscular force of its occupant (s) and which is not equipped with an engine, such as a conventional scooter, a bicycle, skis / skates / boards roller skates, ice skates, snow skis, etc.

Motorized wheeled vehicle users can be likened to pedestrians as long as they do not use the force provided by an engine to move. For example, the user of a non-motorized scooter performs in a known manner movements back and forth with a leg taking punctually support with the ground to propel itself forward.

According to the invention, this particular movement of the user can also be uniquely characterized by a signature previously recorded in the signature matrix, as previously described. As for the case of the pedestrian, we can distinguish on the basis of the unique signatures different modes of displacement elementary elements according to which the position of the user of the scooter can be determined accurately and autonomously.

Naturally, to meet specific needs, a person skilled in the field of the invention may apply modifications to the invention as described above without departing from the scope of the invention. For example, the motion recognition algorithm may be executed by the application installed on the mobile phone rather than by the application of the data sensor.

The present invention is not limited to the specific embodiments that have been described above, and the modifications that are within the scope of the present invention will be apparent to one skilled in the art.

Claims

A method of autonomous geolocation of a first person moving on foot or by means of a non-motorized moving device, said method intended to be implemented on a portable device integral with said person, said method being characterized in that it includes the following steps:

• measurement (El) of inertial data relating to a displacement of said first person;

Recognition (E3) of a basic mode of movement of said first person, in which all or part of the measured inertial data are compared with predetermined signatures, each signature being previously uniquely associated with an elementary mode of movement independently of said first person considered;

Estimating (E5) the speed and direction of movement of said first person, according to the recognized elementary displacement mode and all or part of the measured inertial data; and

• evaluation (E7) of the position of said first person, according to the estimated speed and direction of movement and the position evaluated during a previous iteration of the evaluation step.

Method according to claim 1, characterized in that it comprises a first preliminary step (E01), in which each signature is elaborated on the basis of at least one temporal signal of inertial data obtained for a given elementary displacement mode and performed by at least one second person.

3. Method according to any one of the preceding claims, characterized in that during the recognition step (E3), the measured inertial data are compared to the signature elements of each of the displacements. predefined elementary elements, by calculating a weight-weighted correlation score (W) previously assigned to each type of inertial data, so that the recognized displacement mode is the one for which the correlation score is the highest.

Method according to any one of the preceding claims, characterized in that during the recognition step (E3), the inertial data are selected from data of vertical acceleration, frontal, transverse, rotation.

Method according to the preceding claim, characterized in that the vertical acceleration data are assigned a multiplicative factor constituting a preponderant weight compared to the other inertial data compared during the recognition step (E3).

Method according to the preceding claim, characterized in that the vertical acceleration data are assigned a first weight and that:

• the frontal acceleration data is assigned a second weight of value at least two times lower than that of the first weight, and / or

• the transverse acceleration data are assigned a third weight of value at least eight times lower than that of the first weight, and / or

• the roll type rotation data are assigned a fourth value weight at least half of that of the first weight, and / or

Pitch-type rotation data are assigned a fifth weight value at least four times less than that of the first weight.

Method according to one of the preceding claims, characterized in that the inertial data compared during the recognition step (E3) correspond to a sample of time data of a duration between 1 s and 2 s.

8. Method according to any one of the preceding claims, characterized in that the speed of displacement of said first person is estimated, during the estimation step (E5), according to the recognized elementary displacement mode and a movement rate of said first person estimated in real time from the measured inertial data.

9. Method according to the preceding claim, characterized in that the evaluation of the displacement speed is carried out from a model previously established and defining for each mode of elementary displacement a unique relationship between the rate and the speed of displacement.

10. Method according to the preceding claim, characterized in that said model is produced by polynomial regression, on the basis of a set of discrete measurement points, each point associating a speed with a rate of displacement.

11. Method according to any one of the preceding claims, characterized in that the estimation of the direction of movement of said first person comprises:

A filter sub-step (E52) in which all or part of the measured inertial data is filtered by subtracting corresponding data included in the signature of the recognized mode of movement; and

A correction sub-step (E54) of a systematic error b (t _n ) inherent in the measurement means (1) of said inertial data.

12. Autonomous geolocation device for a person moving on foot or by means of a non-motorized moving device, said device being intended to be integral with said person, said device being characterized in that it comprises: Measuring means (1; 10, 12) for measuring inertial data relating to a displacement of said person;

Recognition means (4) for recognizing a mode of elementary displacement of said person, said recognition means being adapted to compare all or part of the inertial data measured at predetermined signatures, each signature being previously stored in storage means uniquely associated with a particular mode of movement independently of said subject;

Estimation means (4) for estimating the speed and direction of movement of said person according to the recognized elementary displacement mode and the measured inertial data; and

Evaluation means (4) for evaluating the position of said person according to the estimated speed and direction of movement and the position evaluated during a previous iteration of the evaluation step.

13. Computer program (PGl) comprising instructions adapted to the implementation of at least one of the steps of the method according to any one of claims 1 to 11, when said program (PGl) is executed on a computer .

14. Information storage medium (5), removable or not, partially or completely readable by a computer or a microprocessor comprising code instructions of a computer program (PG1) for the execution of at least one any of the steps of the process according to any one of claims 1 to 11.