WO2018050986A1

WO2018050986A1 - Step detection technique

Info

Publication number: WO2018050986A1
Application number: PCT/FR2017/052328
Authority: WO
Inventors: Julien Faure; Gaétan FROISSARD
Original assignee: Orange
Priority date: 2016-09-13
Filing date: 2017-09-04
Publication date: 2018-03-22
Also published as: FR3055957A1

Abstract

The invention relates to a technique for step detection by a user device. This method comprises: - obtaining (E1) an acceleration signal from successive acceleration measurements along at least one axis in a reference frame of three orthogonal axes, in the form of successive samples; - processing said signal, in the course of which a current sample of the acceleration signal is saved (E5) when the absolute value of its amplitude is greater than a dynamic threshold dependent on a sliding average determined over a plurality of acceleration signal samples centred on said current sample, the saved samples forming a processed signal; - a search (E6-E7) in the processed signal for at least two samples, wherein the first sample takes an extreme value of a first type belonging to the group comprising a minimum and a maximum, and the second sample is the first sample found among the following samples that takes an extreme value of the second type, the second type belonging to said group and being distinct from the first type, a step being detected when said two samples are found.

Description

Step detection technique

The invention relates to the general field of telecommunications.

The invention relates more particularly to a step detection technique performed by a user of a device. This device is integral with the movements of the user and thus the user.

There is currently a craze for devices or software tracking activity for sport or everyday life. These are intended to give an estimate of the activity of their user. One of the activities that can be followed is walking or running represented by movements made over a period of time or on a path and characterized by a number of steps performed.

For example, patent document WO2011095842 describes an inertial device for detecting a step of a user of this device. More specifically, a step is detected from a measurement of an acceleration along three orthogonal axes, performed by a sensor. An effective value, or RMS value (for "Root Mean Square"), is determined from this acceleration measurement. This rms value is then compared with a threshold value, a step being detected when the rms value is greater than this threshold value.

It has been found that such a device is particularly sensitive to the conditions in which the acceleration measurement is made. Poor attachment of the sensor performing the measurement can indeed lead to a decision to detect a step that is not correct. User movements, such as getting up from a chair, can also lead to false detections. Because of these false detections, the device will tend to overestimate the activity.

One of the aims of the invention is to remedy the shortcomings / disadvantages of the state of the art and / or to make improvements thereto.

According to a first aspect, the subject of the invention is a method of detection of steps by a user device, in solidarity with a user. The method comprises:

obtaining an acceleration signal from successive acceleration measurements along at least one axis in a reference frame of three orthogonal axes in the form of successive samples;

a processing of said signal, in which a current sample of the acceleration signal is kept when the absolute value of its amplitude is greater than a dynamic threshold dependent on a sliding average determined on a plurality of samples of the acceleration signal centered on said current sample, the conserved samples forming a processed signal;

a search in the processed signal of at least two samples, in which the first sample takes an extreme value of a first type belonging to the group comprising a minimum and a maximum, and the second sample is the first found among the following samples taking an extreme value of the second type, the second type belonging to said group and being distinct from the first type, a step being detected when said two samples are found.

The device is worn by a user or fixed to a user and is thus integral with the movements of this user.

By detection of a step, subsequently means a detection of an impact on the ground. This impact is produced by the contact of a foot for a human being or a paw for an animal on the ground.

The detection method thus uses a dynamic threshold, which depends directly on the moving average of an accelerometer signal measured by an accelerometer or an acceleration signal that has been pre-processed. It is then possible to detect steps in very different motion conditions: walking, running, climbing a staircase. It is also possible to detect steps in very different measurement conditions: unstable accelerometer fixation, variable location (right or left pocket of trousers, handbag). Moreover, this method makes it possible to detect steps of both humans and animals.

This detection method has the advantage of relying solely on acceleration measurements provided by an accelerometer. It does not require measurements provided by a gyroscope, the latter being very consumer of energy resources. As an illustrative example, the energy consumption of a gyroscope is about a hundred times higher than that of an accelerometer alone.

Thus, the detection method is particularly suitable for being integrated into a connected object, such as a connected watch, a connected bracelet or even a connected collar. It can also be executed on a mobile terminal, such as a "smartphone".

The different modes or features of embodiment mentioned below may be added independently or in combination with each other, to the steps of the detection method as defined above.

According to a particular characteristic of the detection method, a sample of the acceleration signal is obtained by subtracting said sliding average from an acceleration measurement.

Acceleration signal readings have shown that when the average speed, that is the user's rate, varies, the acceleration signal no longer oscillates around a constant mean value. This is particularly the case during an acceleration during a race or during a transition from walking to running. It is then more difficult to detect a sequence of a sample taking a maximum value followed by a sample taking a minimum value. Thanks to this centering operation around the sliding average, the acceleration signal is centered whatever the user's rhythm and the detection of a linking is facilitated. The detection of a step is thus improved when the pace of the user varies.

According to a particular characteristic of the detection method, a low-pass filter is applied to the acceleration measurement prior to the determination of said sliding average.

This low pass filter eliminates noise and suppresses undesirable features in the acceleration signal. This noise and these characteristics are likely to degrade the acceleration signal. They can be introduced by accelerometer based on its performance, a temperature of use, vibration or instantaneous change of pace. Low-pass filtering thus smoothes the acceleration signal. Such a low-pass filter is easily integrable into a real-time program and requires little computing resources. The low-pass filter is for example a filter of order one.

According to a particular characteristic of the detection method, a step is counted when a rolling average of the duration of previous steps is between a first and a second threshold value.

This makes it possible to exclude steps whose values are assimilated to detection errors, for example because they do not correspond to current values observed in humans or animals. For example, the vibrations of a car engine that cause accelerations large enough to be of the same order of magnitude as steps but whose frequency characteristics (number of oscillation per minute) are far removed from the values of not current .

According to a particular characteristic of the detection method, cumulative or alternative to the preceding characteristic, a step is counted when a variance of duration of previous steps is less than a third threshold value.

This makes it possible to discard steps whose values are assimilated to detection errors, because of a too great variability compared to the other steps detected previously. Indeed, over a short observation period of a few steps, we can consider that the user is walking or running at a relatively regular pace.

According to a second aspect, the invention also relates to a user device, intended to be integral with a user, said device comprising:

a module for obtaining an acceleration signal from successive acceleration measurements according to at least one axis in a reference frame of three orthogonal axes in the form of successive samples;

a detection module, arranged for:

processing an acceleration signal obtained by keeping a current sample of the acceleration signal when the absolute value of its amplitude is greater than a dynamic threshold dependent on a sliding average determined on a a plurality of samples of the acceleration signal centered on said current sample, the conserved samples forming a processed signal;

search in the processed signal for at least two samples, in which the first sample takes an extreme value of a first type belonging to the group comprising a minimum and a maximum, and the second sample is the first found among the following samples taking an extreme value of the second type, the second type belonging to said group and being distinct from the first type, a step being detected when said two samples are found. This user device can of course include in structural terms the various characteristics relating to the detection method as described above, which can be combined or taken in isolation. Thus, the advantages stated for the detection method according to the first aspect can be transposed directly to the user device. Therefore, they are not detailed further.

According to a third aspect, the invention relates to a program for a user device, comprising program code instructions for controlling the execution of the steps of the previously described detection method implemented by the user device, when this program is executed. by this device and a recording medium readable by a device on which is recorded a program for a device.

The advantages stated for the detection method according to the first aspect can be transposed directly to the program for a user device and to the recording medium.

The detection technique will be better understood with the following description of particular embodiments, with reference to the accompanying drawings, in which:

FIG. 1 represents an environment in which the detection method is implemented in a particular embodiment;

FIG. 2 represents a user device according to a particular embodiment;

FIG. 3 illustrates steps of a detection method implemented by a user device according to a particular embodiment.

FIG. 1 represents an environment in which the step detection method is implemented during a displacement of a user device in a particular embodiment. The user device is intended to be secured to a user: it can be worn, fixed, ... Thus, the displacement of the user device corresponds to a displacement of the user of which this user device is secured. This movement can be of type walking, running, climbing or descending a staircase, or any combination of these types of displacement. By user, we mean afterwards both a human being and an animal. By detecting a step, hereinafter is meant detection of an impact on the ground. This impact is produced by the contact of a foot for a human being or a paw for an animal on the ground.

The environment shown comprises two user devices 10, 11 accessing a communication network 1 via a mobile access network, not shown in FIG. 1. The mobile access network corresponds, for example, to a network. mobile communication type GSM, EDGE, 3G, 3G + or 4G (also called LTE for "Long Term Evolution") ... In the embodiment shown, the user devices 10, 11 send data relating to an activity of the to a server 20. This server 20 is arranged to provide a service from the received data. This service may correspond for example to a health service, a personal assistance service. It is emphasized here that the step detection technique does not require this access to the communication network 1 to be implemented. The user devices 10, 11 can thus operate completely independently. Moreover, in a particular embodiment, the user devices 10, 11 store the data relating to a user activity in a local storage memory, in order to send them later to the server 20, for example when the communication network 1 is accessible.

The user devices 10 and 11 can be any type of terminal for transmitting data relating to a user's activity, such as a mobile phone, a smartphone, a tablet, a connected object. .

In the example shown in FIG. 1, the user device 10 is a "smartphone" type terminal.

In the example shown in Figure 1, the user device 11 is a connected object having a screen and interaction buttons.

These two user devices each comprise a sensor (or obtain data coming from an external sensor), arranged to measure a linear acceleration of the user device, and therefore of the user, in a reference frame composed of three orthogonal axes denoted x, y, z. More specifically, the internal or external sensor is arranged to measure three linear accelerations, each along one of the orthogonal axes. This sensor is called thereafter three-axis accelerometer. FIG. 2 represents a user device 10, 11 in a particular embodiment. The user device 10, 11 comprises in particular:

a processor 110 for executing software module code instructions;

a memory zone 111, arranged to store an application that includes code instructions for implementing the steps of the detection method;

a storage memory, not shown, arranged to store data used during the implementation of the detection method, data relating to a user activity or else data resulting from the implementation of the method of detection; a module 112 for obtaining an acceleration signal from successive acceleration measurements according to at least one axis in a reference frame of three orthogonal axes in the form of successive samples;

a step detection module 113.

In a particular embodiment, the obtaining module 112 obtains the acceleration signal from an external sensor comprising a three-axis accelerometer. This external sensor is integral with the user device.

In another particular embodiment, the obtaining module 112 is a sensor comprising a three-axis accelerometer, arranged to measure an acceleration 31.25 times per second.

In a particular embodiment, the user device 10, 11 further comprises a transmission / reception module, arranged to communicate with the communication network 1 via a mobile access network.

In a particular embodiment, the user device 10, 11 comprises a low power radio transmission / reception module. As an illustrative example, it is a radio module BLE, for "Bluetooth Low Energy". This allows the user device 10, 11 to transmit its data to another device, for example a mobile terminal, which has access to the communication network 1.

In a particular embodiment, the sensor 112 also comprises a 3-axis gyroscope. However, for the implementation of the detection technique, it is not necessary to activate it, which limits the energy consumption.

In a particular embodiment, the user device 10, 11 comprises a user interface module.

It is emphasized here that for the implementation of the detection method, it is sufficient that a measurement of the acceleration on an axis is possible.

It is also pointed out that the user device 10, 11 also comprises other processing modules, not shown in FIG. 2, arranged to implement the different user device functions. FIG. 3 illustrates steps of a step detection method implemented by a user device 10, 11 according to a particular embodiment.

It will be recalled here that the detection method only uses successive acceleration measurements, independently of the orientation of the user device 10, 11. Thereafter, one places oneself in the particular case where three Acc _x Acceleration measurements, Acc _y , Acc _z are provided, each of which is associated with one of the orthogonal axes x, y, z. It is recalled that the detection method can also be implemented as a function of acceleration measurements made on a single axis. It is placed subsequently in the particular case where the user device 10, 11 comprises the sensor. The description is easily modifiable in the particular case where the obtaining module 112 of the user device 10, 11 obtains the acceleration signal of an external sensor.

In a step E1, the detection module 113 obtains from the sensor 112 an acceleration signal from successive acceleration measurements according to at least one axis in a coordinate system of three orthogonal axes x, y, z in the form of samples. successive Acc _x , Acc _y , Acc _z . The acceleration measurements are obtained with a fixed frequency, for example 31.25 Hz, using an analog digital converter integrated in the sensor 112.

In a step E2, the detection module 113 puts the acceleration signal in the form of a normalized acceleration signal, representing the Euclidean norm of the acceleration measurement according to each of the three axes.

where Acc _N represents the normalized vector of acceleration, the Acc "Acc _y , Acc _{z components} correspond respectively to the acceleration measurements along the x, y and z orthogonal axes of the sensor 112 and n is the index of the sample .

It was found by the applicant from the acceleration signals obtained that the signals corresponding to the different axes of the accelerometer Acc _x , Acc _y , Acc _z vary almost identically and that only the amplitudes of the signals change. It is thus possible in a particular embodiment to perform a step detection with a single acceleration measurement obtained for a single axis of the sensor. This embodiment is particularly suitable when the orientation of the sensor is defined and does not vary.

In a step E3, the detection module 113 applies a low-pass filter, for example of the first order, to the normalized acceleration signal Acc _N , representative of the measurement of acceleration or of the acceleration measurements for the sample. not.

More precisely :

Acc ' _N (n) = Acc' _N (n-1) - (Acc ' _N (n-1) - Acc _N (n)) where a is the coefficient of the filter, Acc' _N is the filter output and Acc _N corresponds to the normalized acceleration signal. We can set for example a to 1/20.

This low-pass filter has the effect of smoothing the curve and limiting the noise in the acceleration signal. This allows in particular to improve the performance during the detection of extrema which is the subject of a step described later. It is emphasized here that such a low-pass filter is easily integrable in a real-time program and requires little computing resources.

In a step E4, the detection module 113 determines a sliding average of the normalized acceleration signal Acc ' _N obtained during the step E3 and subtracts this sliding average determined at sample n of the normalized acceleration signal. This sliding average corresponds locally to the continuous component of the signal.

This sliding average M is determined by the detection module 113 on a K sample window centered on the sample n. We denote K in the form 2L + 1.

More precisely :

M (n) = -

where L is a constant.

As an illustrative example, one can set K to the value of seventeen and so the constant L takes the value of eight.

For the sample n, the new acceleration signal S obtained then corresponds to:

5 (n) = Acc ' _N (n) - (n)

The centering of the acceleration signal as a function of the sliding average makes it possible to improve the detection of extrema, described later. Indeed, it has been observed by the applicant that, when the average speed or the user's rhythm varies, the acceleration signal no longer oscillates around a constant mean value. This complicates the detection of extrema. The oscillations of the acceleration signal are thus reduced around the zero value, irrespective of the variations in the speed of the user.

In a step E5, the acceleration signal S is processed by the detection module 113 in order to keep only the samples whose amplitude is greater than a dynamic threshold. More specifically, a current sample of the acceleration signal S is maintained when the absolute value of its amplitude is greater than a dynamic threshold dependent on a moving average M (n) determined on a plurality of samples of the centered acceleration signal. on the current sample of index n. The preserved samples form a processed acceleration signal S '.

This treatment corresponds to:

s, _{(n) =} (5 (n) 5i | 5 (n) |>? (n) / G

0 otherwise

where β is a constant and G is the value of gravity 10 ms ^"2 at the output of the digital-to-digital converter of the sensor 112.

The constant β for example takes the value of one hundred.

For a converter delivering acceleration measurements encoded on 16 bits, varying between plus or minus 4 times G, then the value of 1 G is encoded by 8192.

Since the dynamic threshold depends on the moving average of the acceleration signal, the higher the user 's rate, the greater the noise suppression. This then makes it possible to better detect extrema. The dynamic threshold makes it possible to adapt to changes in the user's rhythm.

In the embodiment described, the detection module 113 then searches the processed acceleration signal S 'for at least one sample taking a maximum value followed by the following samples of a sample taking a minimum value. By tracking means that in the processed acceleration signal there is a sample taking a maximum value and a sample taking a minimum value, succeeding in time without necessarily being contiguous. No other sample taking extreme value is found between the two samples found. A step is detected when such a sequence is found. It is emphasized here that in another particular embodiment, a step is detected by a sequence of a sample taking a minimum value followed by a sample taking a maximum value. In other words, the detection module 113 searches in the processed acceleration signal S 'for at least two samples, in which the first sample takes an extreme value of a first type belonging to the group comprising a minimum and a maximum, and the second sample is the first found among the following samples taking an extreme value of the second type, the second type belonging to said group and being distinct from the first type, a step being detected when said two samples are found.

A maximum value is detected if the difference between two successive samples is positive then negative. In other words, when {S '(n) - S' (n-1)} is strictly positive and {S '(n + 1) - S' (n)} is strictly negative, then a maximum value has been detected. In the described embodiment, the processed acceleration signal has been centered: a maximum value is always positive and the non - positive maximum values are discarded.

A minimum value is detected if the difference between two successive samples is negative then positive. In other words, when {S '(n) - S' (n-1)} is strictly negative and {S '(n + 1) - S' (n)} is strictly positive, then a minimal value has been detected. In the embodiment described, the processed acceleration signal has been centered: a minimum value is always negative and non - negative minimum values are discarded.

In a particular embodiment, this search is implemented as follows.

In a step E6, the detection module 113 calculates the difference between {S '(n + 1) -

S '(n)} and compare its sign to that of {S' (n) - S '(n-1)}. If the sign is identical, then S '(n) does not correspond to an extreme value (minimum or maximum) in the processed acceleration signal. The detection module 113 then returns to the step E1 waiting to receive a new sample.

In the opposite case, that is, if the sign of {S '(n + 1) - S' (n)} is different from that of

{S '(n) - S' (nl)}, a minimum or maximum value is detected. In a step E7, the detection module 113 stores in a memory zone the detection of this value. Always in this step E7, when the detected value is minimal and there has been a detection of a maximum value among the preceding samples, the detection module 113 also stores the index of the sample n which led to the detection of this minimum value, a step p having been detected. Thus, in this embodiment, the duration of a step corresponds to the difference in terms of samples between two minimum value detections.

D (p) is a vector of durations (in terms of samples) between two successive steps p determined by the implementation of steps E1 to E7. D (p) thus corresponds to the time elapsed between step p and step p + 1.

In a step E8, the detection module 113 filters on the step detections, in order to eliminate false detections.

More precisely, in this step E8, the detection module 113 determines a sliding average D 'of step duration and a pseudo-variance Var of step duration. In the embodiment described, these mean and pseudo-variance are determined as follows:

Var (p) = ¹ / _p \ D (i) - D (p) \

* Y-i ⁱ i = ^P pPi

where P is a constant.

For example, P can take a value of eight. This implies that the detection method requires the user of the device to make at least eight steps. Thus, movements less than eight steps will not be accounted for by the process. However, this does not seem very binding when the steps are counted over a significant period.

Thus, the sliding average D 'and the pseudo variance Var are calculated over eight step durations.

In step E8, a step p is taken into account as a step when the running average D 'of previous step duration is between a first threshold value C1 and a second threshold value C2. It is assumed that a user moves at a speed between 0.5 and 5 steps per second. We can therefore choose Cl at 6 and C2 at 62. When the sliding average is not between these two thresholds, then it is probably a question of parasitic noise, for example linked to a shock during the rise of a stair. In this case, the step p is not taken into account and is not finally detected; the detection method returns to step El, waiting to obtain an acceleration signal from the sensor 112.

Still in this step E8, a step p is taken into account when the variance of duration of previous steps is less than a third threshold value C3. More precisely, the pseudo-variance Var as determined previously is compared with the third threshold value C3. It is assumed here that the user walks or runs at a relatively regular rate. We choose example C3 at the value 15. When the variance of duration of previous steps is greater than the third threshold value C3, then it is then probably noise. In this case, the step p is not taken into account and is not finally detected; the detection method returns to step El, waiting to obtain an acceleration signal from the sensor 112.

So, a step p is finally detected during step E8 when the running average

D (p) of previous step duration is between the first threshold value C1 and the second threshold value C2 and when the variance Var (p) of previous step duration is less than the third threshold value C3:

(true if Cl≤D '(v) ≤ C2 and Var p) <C3

not = ■

(fake otherwise

In a step E9, the step p is counted. A counter C _step is incremented and represents a number of steps detected over an observation period. The detection method returns to step E1, waiting to obtain an acceleration signal from the sensor

112.

The embodiment which has been described makes it possible to determine a number of steps performed by a user of a user device reliably and only from a measurement of an acceleration signal. This limits the energy consumption of the device. The performance of the detection method proved to be better than other known algorithms, especially when tested in different environments with a panel of users. This detection method is easily integrable into a connected object.

In a particular embodiment, the calculations during the implementation of steps E2 to E8 are performed in integer values. This makes it possible to reduce the complexity of implementing the method.

No limitation is attached to these various embodiments and the skilled person is able to define others based on a dynamic threshold dependent on a sliding average determined on a plurality of samples of a signal acceleration centered on a current sample.

In a particular embodiment, the step E4 of centering the acceleration signal around the sliding average is not implemented. In this case, the step E5 is adapted with a dynamic threshold which always depends on the sliding average but taking into account this lack of centering.

In a particular embodiment that can be cumulative with the previous embodiment, the step E3 of low-pass filtering is not implemented. The performance of the process remains satisfactory.

In a particular embodiment that can be cumulative with the previous embodiments, step E8 for filtering step times is not implemented. The performance of the process remains satisfactory. In the embodiment that has been described, the step E8 is implemented with two comparisons, one on the average step duration and the other on the variance of the step duration. Other embodiments may also be provided by performing only one of the two comparisons.

The detection technique is implemented by means of software and / or hardware components. In this context, the term "module" may correspond in this document to both a software component, a hardware component or a set of hardware and / or software components, capable of implementing a function or a set of functions, as described above for the module concerned.

A software component corresponds to one or more computer programs, one or more subroutines of a program, or more generally to any element of a program or software. Such a software component is stored in memory and then loaded and executed by a data processor of a physical entity and is able to access the hardware resources of this physical entity (memories, recording media, communication buses, electronic cards of a physical entity). input / output, user interfaces, etc.).

In the same way, a material component corresponds to any element of a material set (or hardware). It may be a programmable hardware component or not, with or without an integrated processor for running software. This is for example an integrated circuit, a smart card, an electronic card for executing a firmware, etc.

In a particular embodiment, the modules 112, 113 are arranged to implement the previously described detection method. These are preferably software modules comprising software instructions for performing the steps of the detection method described above, implemented by a user device. The invention therefore also relates to:

a program for a user device, comprising program code instructions for controlling the execution of the steps of the previously described detection method, when said program is executed by this user device;

a recording medium readable by a user device on which the program for a device is recorded.

Claims

A method of step detection by a user device (10, 11), integral with a user, said method comprising:

obtaining (El) an acceleration signal from successive acceleration measurements according to at least one axis in a coordinate system of three orthogonal axes in the form of successive samples;

a processing of said signal, in which a current sample of the acceleration signal is kept (E5) when the absolute value of its amplitude is greater than a dynamic threshold dependent on a sliding average determined on a plurality of samples of the signal acceleration centering on said current sample, the conserved samples forming a processed signal;

a search (E6-E7) in the processed signal of at least two samples, in which the first sample takes an extreme value of a first type belonging to the group comprising a minimum and a maximum, and the second sample is the first found among the following samples taking an extreme value of the second type, the second type belonging to said group and being distinct from the first type, a step being detected when said two samples are found.

The detection method according to claim 1, wherein a sample of the acceleration signal is obtained (E4) by subtracting said moving average from an acceleration measurement.

3. Detection method according to claim 2, wherein a low-pass filter is applied (E3) to the acceleration measurement prior to the determination of said sliding average.

4. Detection method according to any one of claims 1 to 3, wherein a detected step is counted (El i) when a rolling average of previous step duration is between a first and a second threshold values.

5. Detection method according to any one of claims 1 to 4, wherein a detected step is counted when a variance of previous step duration is less than a third threshold value.

User device (10, 11), intended to be integral with a user, said device comprising:

- A module for obtaining (112) an acceleration signal from successive acceleration measurements along at least one axis in a reference of three orthogonal axes in the form of successive samples; a detection module (113), arranged for:

processing an acceleration signal obtained by keeping a current sample of the acceleration signal when the absolute value of its amplitude is greater than a dynamic threshold dependent on a sliding average determined on a plurality of samples of the acceleration signal centered on said current sample, the conserved samples forming a processed signal;

search in the processed signal for at least two samples, in which the first sample takes an extreme value of a first type belonging to the group comprising a minimum and a maximum, and the second sample is the first found among the following samples taking an extreme value of the second type, the second type belonging to said group and being distinct from the first type, a step being detected when said two samples are found.

7. Program for a user device, comprising program code instructions for controlling the execution of the steps of the detection method according to one of claims 1 to 5 implemented by the device, when said program is executed by said device.

8. Recording medium readable by a user device on which the program according to claim 7 is recorded.