EP4330838A1 - Characterisation of a user via association of a sound with an interactive element - Google Patents

Abstract

The invention relates to a method for characterising a user and to a corresponding computer program, storage medium and processing device. The method comprises a comparison (S6) between at least a first sub-set of first identifiers of a first sound and at least one second identifier received from a human-machine interface, while a second sound is being listened to by a user of the human-machine interface. The comparison triggers, if the at least one received second identifier belongs to the first sub-set associated with a first sound corresponding to the expected second sound, a characterisation of the user as being a suitable user.

Description

Description

Title: Characterization of a user by association of a sound with an interactive element

Technical area

[0001] This disclosure relates to the field of computer security.

More particularly, the present disclosure relates to a method for differentiation between a human and a machine and to a corresponding computer program and recording medium.

Prior technique

[0003] A CAPTCHA ("Completely Automated Public Turing-test to tell Computers and Humans Apart") is a test intended to differentiate a user from a program computer science. It is used in particular on forms to protect against spam, that is to say unsolicited electronic communications.

[0004] There are different types of CAPTCHAs, the most numerous are visual CAPTCHAs where the user is asked to enter a series of distorted letters displayed on the screen.

[0005] An accentuated deformation is essential in terms of computer security. Indeed, image recognition being very widespread, it is increasingly easy for a computer robot to hijack this kind of security bricks.

[0006] In return for such accentuated distortion, CAPTCHAs often pose difficulties even for the most seasoned users. Indeed, it is often necessary to make several attempts before responding correctly to a CAPTCHA.

[0007] For some users, the presence of a CAPTCHA is simply prohibitive. For example, a blind user cannot fill in a visual CAPTCHA. These verification systems fail to recognize some users with disabilities as human and therefore make it impossible for these users, to create accounts, to write comments or to make purchases on certain sites.

[0008] Some sites offer, as an alternative for users who so wish, an audio CAPTCHA in which a voice recites a sequence of numbers to be transcribed in the user's supposed language.

[0009] Nevertheless, the audio description is conditioned on the language of the user and his level of understanding of the language used, even native in cases of illiteracy. Thus, audio description remains a barrier to accessibility for a certain number of users.

[0010]There is therefore a need for a new method of man-machine differentiation that overcomes language barriers, for the purposes of increased accessibility.

Summary

[0011] This disclosure improves the situation.

[0012] A method for characterizing a user is proposed, the method being implemented by a processing device, the method comprising:

- a comparison between at least a first identifier of a first sound and at least a second identifier received from a man-machine interface, the second identifier being received after a reproduction of a plurality of second sounds by the man-machine interface machine, the comparison triggering, if the second identifier received corresponding to the first identifier, a characterization of the user as being an appropriate user.

[0013] The term "processing device" means any device capable of performing computer processing according to the proposed method. It may be, for example, a terminal for use by a user such as a television set, a computer or a multifunction telephone, or network equipment, in particular a home gateway, a set-top box, or any form of device for providing digital services and/or digital content, accessible for example via a local network and/or a wide area network. [0014] It is possible, for example, to define several first sounds, such as a drum roll, a kettle whistle, thunder, etc., and associate, with each of these first sounds, one or more first identifiers. Thus, for each of these first sounds, a first corresponding subset is obtained, composed of one or more first identifiers. The first sound corresponding to a drum roll can thus be associated with various identifiers such as for example a word “drum”, an expression “drum roll”, different images of drums, possibly animated, etc.

[0015] It is considered that the second sound, heard by the user, corresponds to one of the first sounds. For example, the second sound may include a drum roll identical or similar to that constituting one of the examples of first sounds defined above. The user then interacts with the man-machine interface so as to associate at least a second identifier with the second sound.

The method thus makes it possible, thanks to the comparison between at least a second identifier and a first identifier, to determine whether the identification by the user of the second sound heard is consistent with the expected identification. This makes it possible to certify, in the event of success, that the user is able to correctly identify the second sound and to conclude by characterizing the user as being an appropriate user. Conversely, this also makes it possible to characterize a user who has not succeeded in correctly identifying the second sound as being an inappropriate user, or at least not to characterize him as being an appropriate user.

In a first particular embodiment, the user will identify one of the second sounds heard prior to a question relating to one of these second sounds heard, for example, after having reproduced 5 animal cries, the following question is asked to the user: "What animal is heard third in the list of calls?" and the second identifier supplied/selected by the user for the second sound heard in the third position of the list of second sounds reproduced will be compared with the first identifier stored in association with a first sound corresponding to the second sound heard in the third position of the list second sounds reproduced. In a second particular embodiment, the user will identify one of the second sounds heard prior to a question relating to one of these second sounds heard, for example, after having successively reproduced 5 second sounds (each sound possibly being associated with a graphic element such as a box in a mosaic), the following question is asked to the user: "In which position did you hear the or which box corresponds to the animal cry? and the second identifier provided/selected by the user for the second sound heard will be compared to the first identifier, namely in this case the position/the identifier of the box, associated with a first sound associated with the corresponding second sound heard to the animal cry from the list of second sounds reproduced.

In a third particular embodiment, the user will identify each of the second sounds heard in the order of reproduction of these second sounds heard and each of the second identifiers provided/selected by the user for each second sound heard will be compared to a first identifier stored in association with a first sound corresponding to each second sound heard from the list of second sounds reproduced.

[0020] A sound captcha according to the invention is a sound "memory" since the user must memorize each sound heard in order to then identify one or more sounds according to the order of reproduction which is also the order listening.

[0021] The expression “appropriate user” can designate, for example, a human capable of correctly identifying the second sound as opposed to a machine devoid of such a capacity. Alternatively, the expression “appropriate user” can designate a user equipped with specific equipment, such as an audio output capable of broadcasting the second sound to allow the user to hear it, or such as an audio input capable of capturing the second sound as it is heard by the user, as opposed to a user not equipped with such hardware.

[0022] A first example of variation of the above method is proposed. In this first example of variation, the method comprises:

- a command to broadcast a sound extract, - a command for displaying a set of thumbnails, the set of thumbnails displayed comprising a first subset formed of at least one thumbnail, the first subset being associated with the sound extract broadcast,

- reception of at least one selection signal from a man-machine interface, a signal being indicative of a selection of a thumbnail from the set of thumbnails displayed,

- on the basis of the at least one selection signal received, an identification of a second subset formed of each selected thumbnail,

- a comparison test between the first subset and the second subset, and

- based on the test result, an identification of a user as a human or as a machine.

[0023] A second example of variation of the above process is also proposed, in which the process comprises:

- a command for displaying a set of playback buttons, one button being respectively associated with a sound extract,

- for each button displayed, upon receipt of an interaction signal, from a man-machine interface, indicative of an interaction with said button, a command to broadcast the sound clip associated with said button,

- after broadcast, for each button, of the sound extract associated with said button:

- a command for displaying a descriptive visual entity of a subset of sound extracts made up of at least one sound extract from among the sound extracts broadcast,

- a command for displaying a set of thumbnails, a thumbnail being superimposed respectively on a play button, the set of thumbnails displayed comprising a first subset of thumbnails formed of at least one thumbnail, the first subset -set of thumbnails being superimposed on a subset of buttons associated with said subset of sound extracts, - reception of at least one selection signal from a man-machine interface, a signal being indicative of a selection of a thumbnail from the set of thumbnails displayed,

- on the basis of the at least one signal received, an identification of a second subset formed of each selected thumbnail,

- a comparison test between the first subset and the second subset, and

- based on the test result, an identification of a user as a human or as a machine.

[0024]Thus, the user performs, in the first example of variation of the method, an association between a sound, that is to say the sound extract broadcast, and at least one image, that is to say say at least one of the thumbnails displayed, by interaction with this thumbnail or thumbnails.

In the second example of variation of the method, the general principle is the same. The user makes an association between an image, the one displayed, and at least one sound, that is to say at least one of the sound extracts broadcast, by interaction with at least one of the thumbnails displayed.

In the two examples of variations of the method, the comparison test makes it possible to verify whether the association made by the user corresponds to an association actually expected, characteristic of an association that a human would make. If so, then the user can be identified as a human, otherwise the user is identified as a machine.

[0027] Insofar as, by the nature of the sound extract(s) broadcast, such as noises or sound effects, and of the image or thumbnails displayed, such a sound-image association is not attached to a particular language , and does not require any particular cultural knowledge, the two variations above of the proposed method both make it possible to differentiate a human from a machine by overcoming the barriers of language and culture, thus allowing better accessibility than known human-machine differentiation methods. [0028] Furthermore, since the sound recognition algorithms are currently less advanced than those for image recognition, this process of human-machine differentiation is also more difficult to circumvent by a malicious actor. Computer security is therefore increased.

[0029] However, the nature of the first and second identifiers is not restricted to an image or a thumbnail.

[0030] Each first identifier and each second identifier can for example have one type of identifier from among the following:

- at least one word, i.e. a word, an expression, a definition, a short text, etc.,

- at least one image, i.e. a photograph, a drawing, an animated image, a video, etc.,

- a label associated with an object; such as a barcode, QR code, FNC tag, etc. allowing the user to select the second identifier in a real or virtual environment,

- a thumbnail.

[0031] Thus, the user generally makes an association between the second sound and a second identifier of any type (image, text, or label in particular) via a man-machine interface. This second identifier is then compared with a subset of first identifiers for the purposes of characterizing the user as appropriate or not.

[0032] A computer program is also proposed comprising instructions for implementing the above method when this program is executed by a processor.

[0033] A non-transitory recording medium readable by a computer on which the above program is recorded is also proposed.

[0034] Such a recording medium can be any entity or device capable of storing the program. For example, the medium may comprise a storage means, such as a ROM, for example a CD ROM or a microelectronic circuit ROM, or even a magnetic recording means, for example a USB key or a hard disk. . [0035] On the other hand, such a recording medium can be a transmissible medium such as an electrical or optical signal, which can be conveyed via an electrical or optical cable, by radio or by other means, so that the computer program it contains can be executed remotely. The program according to the invention can in particular be downloaded onto a network, for example the Internet network.

Alternatively, the recording 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 aforementioned management method.

[0037] There is also proposed a processing device capable of implementing characterization processing of a user on command from a man-machine interface, in which the processing device comprises:

- a comparator able to compare at least a first identifier of a first sound and at least a second identifier received from a man-machine interface, the second identifier being received after a reproduction of a plurality of second sounds by the interface man-machine, the comparator triggering, if the second identifier received corresponding to the first identifier, a characterizer capable of characterizing the user as being an appropriate user.

The above method may optionally include certain additional functions as defined below.

[0039] For example, the method may comprise reception of at least one signal originating from a man-machine interface, the signal comprising the second identifier of the second sound heard.

[0040] The signal may be a voice recording of the user reading, or naming, a second identifier in the form of text.

The signal can result from a tactile interaction of the user with a button on a screen, this button representing the second identifier in the form of a text or an image.

The signal may be the result of an analysis of an acquired image displaying the second identifier in the form of a label associated with an object, this label that can be real, for example a QR code, or virtual, such as an augmented reality tooltip.

[0043] For example, the method may comprise a command to reproduce a sound extract comprising a first sound associated with one of the first subsets of first identifiers of first sounds, the second sound heard comprising the first sound reproduced.

In this example, the comparison between the first subset of first identifiers of the first sound and the second identifier received from the man-machine interface is intended to assess whether a user has heard and correctly identified the first sound. , Already known.

For example, the method may include a command to capture the second sound heard, the capture triggering a comparison of the second sound heard with at least one first sound associated with a first subset of identifiers.

In this example, the comparison between the first subset of first identifiers of the first sound and the second identifier received from the man-machine interface is intended to assess whether the second sound, potentially unknown, finds a match. with a first sound, known, within a database of first sounds.

For example, the method may include a display command for a set of thumbnails, the set of thumbnails displayed comprising at least one identifier of a first subset associated with the second sound heard. This is particularly the case in the first variation above.

[0048] For example, the method may include:

- on the basis of at least one selection signal received comprising at least one second identifier received from a man-machine interface, during the hearing of a second sound by the user of the man-machine interface, a generation of a second subset comprising the second identifiers,

- the comparison being carried out between at least a first subset of first identifiers of a first sound and at least a second identifier of the second subset, the comparison triggering, if at least a second identifier or all the second identifiers of the second subset received belong to the first subset associated with a first sound corresponding to the second sound heard, characterizing the user as being an appropriate user.

For example, the user can thus provide a beam of information identifying the second sound heard, this beam being compared with a beam of reference information identifying a first known sound similar to the second sound heard.

The comparison may comprise a determination of a resemblance metric, for example 100% resemblance if the second identifiers of the second subset are strictly identical to the first identifiers of the first subset and 0% resemblance if none of the second identifiers of the second subset are also found in the first subset and vice versa. This resemblance metric can for example be compared to a threshold beyond which the user is characterized as being an appropriate user. This allows the user some leeway in determining the second identifiers, instead of simply considering the user to be inappropriate if the user's identification of the second sound differs slightly. of the expected identification.

For example, the method can comprise a selection of said at least one first subset of first identifiers in a base of first identifiers on the basis of a correspondence table associating, with each first identifier of the base, a class respectively.

[0052] For example, it may be provided that each subset of first identifiers belongs to a different class. Thus, several subsets of first identifiers can be selected, each being associated with a corresponding first sound, the first sounds in question being of a different type (ringing, roaring, whistling, etc.), or being associated with a different theme ( vehicle, everyday object, animal, ambient noise, etc.). The notion of sound classes is also described below using various examples, a class of first identifiers that can be associated with such a class of sounds.

For example, provision may be made for the selection of said at least one first subset of first identifiers from the base of first identifiers to be further based on the class associated with the sound clip selected as the second sound.

This makes it possible, for example, to ensure that the at least one second identifier which is expected to be associated with the second sound resembles only one of the subsets of first selected identifiers. The robustness of the determination of the user as being or not being an appropriate user is thereby reinforced.

For example, it can be provided that, the second sound being a sound produced by a given element, each first identifier of the first subset of first identifiers associated with the first sound corresponding to the second sound represents the given element.

Many types of given elements can be considered as possible sources of the second sound.

[0057]Thus, the second sound may for example be an animal cry, or an operating noise of a device, or even a noise resulting from a meteorological event.

[0058] Such sound clips have the common advantage of having an easily representable source, in an unambiguous manner for a human being, in visual form.

Thus, for example, provision can be made for the second sound being an animal cry, each identifier of a first subset of first identifiers being associated with the corresponding animal.

[0060] Alternatively, provision can be made that, the second sound being an operating noise of a device, each identifier of a first subset of first identifiers is associated with the corresponding device. [0061] Alternatively, provision can be made that, the second sound being a noise resulting from a meteorological event, each identifier of a first subset of first identifiers is associated with the corresponding meteorological event.

For example, the method can include a selection of the second sound in a base of sound extracts on the basis of a correspondence table associating, with each sound extract of the base, a respective class.

[0063] Classes of sounds can be defined, for example, by grouping together sounds having a source of the same type, for example the noise of rain and the noise of the wind can be grouped together as noises coming from nature, while that a meow and a bark can be grouped together as pet calls. The formation of such classes of sounds, or of noises or sound effects, is known to those skilled in the art.

[0064] Alternatively, classes of sounds can be defined by grouping together sounds having similarities, for example in terms of frequency spectrum.

[0065] For example, the first variation of the above method may comprise a selection of the set of thumbnails in a thumbnail database on the basis of a correspondence table associating, with each thumbnail of the database, a respective class .

[0066] By transposing this principle to the second variation of the method, the latter can analogously comprise a selection of the image in an image database on the basis of a correspondence table associating, with each image of the base, a respective class.

[0067] Classes of images or thumbnails can be defined so that a class corresponds to the same entity represented, or an entity of the same type. For example, a class of images can be formed of images representing an animal, and a subclass of this class of images can be formed solely of images representing a horse. The formation of such classes of images or thumbnails is known to those skilled in the art. The selection of the set of thumbnails in a thumbnail database, or of the image in an image database, can also be performed on the basis of the class associated with at least one selected sound extract.

By referring for example to the first variation of the method, it is possible to select a sound extract belonging to a given class, for example a vacuum cleaner noise belonging to a class of noises from household appliances. On this basis, it is possible to select one, or several, vignettes representing a vacuum cleaner and provide that the other vignettes of the set of vignettes must be selected from classes of vignettes other than that representing household appliances.

This makes it possible to form the set of vignettes by selection from different classes of vignettes and thus to guarantee that no vignette represents an entity similar to that to be identified.

[0071] In this way, ambiguities are avoided that risk causing a human to associate a vignette with a broadcast sound extract which, although objectively legitimate, turns out to be different from the expected association. The relevance of the process to differentiate a human from a machine is reinforced.

[0072] For example, in the first version of the method, the set of thumbnails may also comprise a third subset formed of at least one thumbnail, and said method may comprise:

- a comparison test between the third subset and the second subset, and

- on the basis of this test, an update, in the correspondence table associating, with each sound extract from the base of sound extracts, a respective class, of the class associated with the selected sound extract, or an update updated, in the correspondence table associating, with each thumbnail of the thumbnail base, a respective class, of the class associated with at least one thumbnail of the third subset.

The third subset may for example be formed of one or more thumbnails of the same class as that of the first subset. So Generally, the third subset can represent an entity capable of emitting a noise close to that emitted by the entity represented by the thumbnail(s) of the first subset.

For example, the first subset can represent a kettle, the third subset can represent a roof window in the rain. The hissing of the kettle and the impacts of the rain on the roof window have similarities. In this way, by broadcasting a sound extract which it is not certain whether it corresponds to the whistling of a kettle or the impact of rain on the roof window, it is possible to rely on the association made by the user enters this sound extract and the thumbnails presented in order to be able to characterize this sound extract.

This principle can be applied analogously in the second version of the method, by displaying an image and selecting different sound extracts on the basis of this image, so as to have a first group and a second group. disjoint sound clips, each having a reasonable probability of corresponding to noises emitted by the entity depicted in the image, and a third group of sound clips clearly unrelated to the entity depicted in the image . Thus, the user can associate the first group or the second group of sound extracts with the displayed image and thus be identified as being a human while at the same time offering the possibility of characterizing the group of sound extracts associated with the picture displayed.

Brief description of the drawings

[0076] Other characteristics, details and advantages will appear on reading the detailed description below, and on analyzing the appended drawings, in which:

Fig. 1

[0077] [Fig. 1] illustrates an example of a graphical interface as part of the implementation of a process for differentiating between a human and a machine.

Fig. 2

[0078] [Fig. 2] illustrates in an exemplary embodiment, a variant of the graphic interface of FIG. Fig. 3

[0079] [Fig. 3] illustrates, in an exemplary embodiment, a general algorithm of a computer program for implementing a method for differentiating between a human and a machine.

Fig. 4

[0080] [Fig. 4] illustrates, in an exemplary embodiment, another variant of the graphic interface of figure 1 .

Fig. 5

[0081] [Fig. 5] illustrates, in an exemplary embodiment, a variant of the algorithm of figure 3.

Fig. 6

[0082] [Fig. 6] illustrates, in an exemplary embodiment, a system for implementing a method for differentiating between a human and a machine.

Fig. 7

[0083] [Fig. 7] illustrates, in an exemplary embodiment, another variant of the graphic interface of figure 1 .

Description of embodiments

[0084] The drawings and the description below contain, for the most part, certain elements. They may therefore not only serve to better understand this disclosure, but also contribute to its definition, where applicable.

The general principle of the invention is based on the identification, by a user, of a sound heard. This identification made by the user is then compared to an expected result.

[0086] Specifically,

- the expected result takes the form of a first subset of first identifiers of a “first sound” corresponding to the sound heard, the sound heard being referred to below as a “second sound”, and - The identification made by the user when hearing the second sound takes the form of at least one second identifier received from a man-machine interface.

[0087] Based on the comparison between the first subset of first identifiers and the second identifier, the user is characterized as being an appropriate user only if the at least one second identifier received belongs to the first subset. Otherwise, the user is characterized as an inappropriate user.

The invention finds many applications in particular in the field of access control to digital services and spam prevention.

For example, the invention can be used, when receiving a message to be published, to characterize the author of the message as being a human or a machine. Thus, the publication of the received message can only be triggered, for example, if the author of the message is characterized as being a human.

For example, the invention can be implemented, when controlling access to digital content having an audio component, to determine whether a user at the origin of a request for access to digital content is equipped with specific equipment, in this case an audio reproduction device such as a headset or earphones. Thus, access to digital content may only be triggered, for example, if the user initiating the request is actually able to listen to the audio component.

In the following description, we endeavor to describe in detail various examples of implementation of the invention.

[0092]0n now refers to Figure 1, which illustrates an example of a graphical interface. A graphical interface is a man-machine interface, that is to say a man-machine dialogue device, in which the objects to be manipulated are drawn in the form of pictograms on a screen, with which the user can interact by means of for example from a pointing device such as a mouse or by digital pressure in the case of a touch screen.

In the example presented here, different sections are represented. [0094] An upper section comprises instructions, in textual form, intended for the user, as well as a button for playing (11) a sound extract comprising a second sound as defined above, and a button volume control (12).

[0095] By “sound excerpt” is meant:

- either an extract from a sound recording, possibly subject to computer processing, including unintelligible audio content,

- or a synthetic sound comprising such audio content.

[0096] Different examples of sounds are presented on https://research.google.com/audioset/ontology/index.html, in the form of several classes of sounds, in this case sounds:

- in connection with the functioning of the human body (voice, breathing, clapping, whistling, heartbeat), or

- emitted by animals (domestic, farm, wild), or

- emitted by objects (vehicle, engine, bell, alarm, tool, explosion, broken wood or glass, controlled flow of liquid), or

- musical (of a specific instrument or musical genre), or

- from nature (wind, storm, fire, rain, waterfall), or

- ambient (noise, acoustic environment), or

- considered ambiguous because they do not formally fall into a single class among those mentioned above (onomatopoeia, impacts, friction).

The sound extracts to which reference is made in the context of the invention are likely to belong to any class of sounds mentioned above, and to any subclass within these classes. , excluding, however, specifically, the intelligible human voice and the sounds attached to a musical genre, insofar as it is desired to overcome linguistic and cultural barriers.

In this case, a first instruction is to listen to the sound extract by interacting with the play button (11). The volume adjustment button (12) is provided to optimize the user's listening comfort. Optionally, an additional button can be provided at the end of reading to allow the sound extract to be listened to again.

[0099] A middle section displays a set (2) of interactive elements, in this case a set of thumbnails (21-29).

[0100] In the example presented here, the vignettes each represent either an animal having a generally recognizable cry or a device having a characteristic operating noise.

A second instruction is to interact with the thumbnail evoking the second sound.

[0102] For example, if the second sound is a neigh, then the user is expected to interact only with the interactive element associated with such a second sound, here the thumbnail representing a horse (28).

[0103] If, on the other hand, the second sound is an engine roar, then it can be expected for example that the user interacts with the interactive element associated with such a second sound, here the thumbnail representing a car (29). Optionally, other interactions can also be accepted. For example, it is conceivable, in this example, that a human interacts alternately with the thumbnail representing a police car (26), or even with that representing an airplane (21).

In the example presented here, the thumbnails are nine in number, distributed in a grid of three by three.

[0105] Such a distribution allows the thumbnails to be displayed easily on different forms of screens, by numerous devices operated by different platforms.

[0106] The number of nine is a compromise between:

- on the one hand, to minimize the probability that a machine interacting with a randomly chosen thumbnail will be incorrectly identified as human, and

- on the other hand, avoid evaluating, in addition to the recognition of sounds, the performance of a human in short-term memory. [0107] It is moreover desirable, still to avoid measuring the performance of the short-term memory of a human, that the thumbnails remain displayed before, during and after listening to the second sound.

[0108] In terms of cognitive psychology, the mnesic span, quantifying the number of digits that can be reproduced immediately after having heard them, usually varies between 5 and 9 depending on the individual. Compared to such a restitution test, a user makes less use of his memory when it comes to selecting, after listening to a sound extract, an interactive element on the basis of its cognitive content. Thus, in the present context, displaying nine thumbnails is acceptable.

[0109] A lower section offers to reset the graphical interface by means of a reset button (31), in order to allow, if the user wishes, to listen to a new second sound then to identify an interactive element corresponding to this new second sound.

We now refer to Figure 2 which is a variant of Figure 1, in which the interactive elements also happen to be thumbnails and in which, after having listened to a second sound, the user is asked to no longer select the interactive element corresponding to this second sound, but all the interactive elements, here all the thumbnails corresponding to this second sound from among a set (4) of interactive elements, here thumbnails (41-49).

For example, the second sound can be an extract from a sound recording of a kettle heating water, or a synthetic sound reproducing such a sound extract.

[0112] Thus, it is possible to define within the set (2) of thumbnails displayed, a first subset of thumbnails, formed of thumbnails (43, 44) each representing a kettle, with each of which an interaction on the part of the user is expected. These thumbnails can thus be associated beforehand, within a correspondence table, with the sound extract indicative of the kettle in operation. More generally, it is possible to define within the set (2) of interactive elements a first subset of elements interactive respectively associated with a first identifier, thus forming a first subset of identifiers associated with a first sound.

It is also possible to define a second subset of thumbnails, formed of the thumbnails with which the user will have actually performed an interaction. More generally, it is possible to define within the set (2) of interactive elements a second subset of interactive elements respectively associated with a second identifier, thus forming a second subset of associated identifiers, by user when hearing a second sound, to this second sound.

[O114JSI the second subset of thumbnails includes all thumbnails from the first subset, then this means that the user has successfully interacted with all the thumbnails associated with the sound bite indicative of the kettle in operation.

[0115] Among the thumbnails displayed that are not part of the first subset, it is possible that some are previously associated, within a correspondence table, with a sound extract different from that indicative of the kettle in operation. For example, the thumbnail (49) at the bottom right of the set of thumbnails shown in Figure 2 may be associated with a sound sample indicative of thunder.

It is also possible that other vignettes, forming a third subset of vignettes, are not previously associated, within a correspondence table, with any sound extract.

[0117] For example, a secondary objective to the characterization of a user as being appropriate or not, for example to the differentiation between a human and a machine, can be to qualify a given sound extract.

[0118] In particular, the sound of rain on a roof window bears a resemblance to the sound of a kettle in operation. To remove the doubt on the nature of a given second sound that could a priori correspond to one or the other of these two possibilities, one can present to a user at the same time:

- two interactive elements, for example two vignettes (43, 44), as a first subset of vignettes, each representing a kettle, and - an interactive element, for example a thumbnail (42), as a third subset of thumbnails, representing a roof window in the rain.

By comparison between the second subset formed of the interactive element(s) with which the user will have interacted and, respectively, the first and the third subset, it is possible:

- on the one hand, to identify the user as being a human, consequently an appropriate user, if the second subset is identical to the first or the third subset, or as being a machine, consequently a user not appropriate, if the second subset is neither identical to the first nor to the third subset, and

- on the other hand, to qualify the sound extract as relating either to the sound of the rain on the roof window, or to the operation of the kettle, in accordance with the interpretation of the user. A correspondence table between sound clips and vignettes can thus be updated to reflect this interpretation.

We now refer to Figure 3 which illustrates a general algorithm of a computer program, for example in connection with one or other of the graphical interfaces of Figures 1 and 2, for the implementation of a method for differentiating between a human and a machine.

By considering a client-server architecture, this computer program can be executed, for example, by one or more processing circuits on the server side, while the interaction with the user takes place on the client side.

In other words, the display of the graphical interface takes place at the level of a terminal on the client side, this terminal possibly being for example a multifunction telephone or "smartphone", a laptop or desktop computer, or a set-top box connected to a television, or network equipment for the provision of services and/or digital content. Similarly, the user can interact, still on the client side, with the graphical interface by means for example of a touch interface, a remote control, a pointing device. On the other hand, the computer program described below can be implemented exclusively on the server side, with the aim of preventing attempts to circumvent the man-machine differentiation. A sound extract, as defined above, is selected (S1) from a database of sound extracts.

Similarly, several interactive elements, here several thumbnails, are selected (S2) from a base of interactive elements, here of thumbnails, thus forming a set of interactive elements, here of thumbnails. For example, 4, 6, 8, 9, 12, 15, or 16 initial thumbnails can be selected to be displayed simultaneously on a 2x2, 2x3, 2x4, 3x3, 3x4, 3x5, or 4x4 grid, respectively.

[0125] One or more additional thumbnails can also be selected with a view to being displayed later, replacing an initial thumbnail having been the subject of an interaction via a man-machine interface.

[0126] The selection of the sound clip and that of the interactive elements are carried out in no particular order. You just need to make sure:

- that one or more of the selected interactive elements represent or are associated with an entity that can be uniquely associated by a human with the selected sound clip, and

- that one or more of the remaining interactive elements represent an entity that can be considered by a human, unequivocally, as unrelated to the selected sound clip.

This objective can be achieved for example by means of statistical learning. Typically, the same second sound can be broadcast to a large number of human users while the same set of thumbnails is simultaneously presented to them. Human users may then be asked to indicate one or more thumbnails corresponding to the sound clip broadcast.

[0128] On the basis of the indications received, it is possible for example to associate one or more mostly selected interactive elements, or one or more mostly selected thumbnails, or selected by a percentage of appropriate users exceeding a certain threshold, with the second broadcast sound. This association can be operated by means of a table of correspondence between, on the one hand, the sound extracts of the base of sound extracts and, on the other hand, the interactive elements of the base of interactive elements.

Thus, it is possible for example to select a given first sound, as well as one or more interactive elements associated with this first sound and forming the first subset of interactive elements associated with a first set of identifiers, and one or more interactive elements not associated with this first sound.

[0130] In addition, it is also possible, for example in the correspondence table defined above, to plan to associate one or more minority-selected interactive elements, or selected by a percentage of appropriate users below a certain threshold. , to an indication that this interactive element, or these interactive elements, are unrelated to this first sound.

Thus, it is possible for example to select a given first sound, as well as one or more interactive elements associated with this first sound and forming the first subset of interactive elements, as well as:

- one or more interactive elements associated with an indication that this or these interactive elements are unrelated to this first sound, and/or

- one or more interactive elements that are neither associated with this first sound nor associated with such an indication.

The first subset of interactive elements is then determined (S3). The first subset of interactive elements is defined as being formed by the interactive element(s) of the set of interactive elements for which an interaction is expected from a user when hearing a second sound corresponding to said first sound.

[0133] Continuing the previous example involving databases and a correspondence table, the first subset of interactive elements designates, in this example, the interactive elements of the set of interactive elements selected which are associated to the first selected sound. A command is issued to trigger the display of a graphical interface such as, for example, that of Figure 1 or that of Figure 2.

Thus, the command triggers the display of the set of interactive elements, here of the selected set of thumbnails (S4) on a display device such as a screen, and can also trigger the display of a play button to start playing the selected sound clip.

The play button is optional, alternatively any other mechanism triggering the playback of the selected sound extract comprising the second sound can be used.

For example, the playback of the sound clip and the display of the set of thumbnails can be triggered, automatically, at the same time. Alternatively, playback of the ssoonnoorree snippet can be triggered automatically after a predefined time expires. This allows the user to prepare, by taking note of the thumbnails as well as the fact that a sound extract is going to be broadcast.

The user is offered to select, from among the set of thumbnails displayed, a thumbnail, or alternatively several thumbnails, corresponding to the selected sound extract.

[0139] Once the selected sound extract has been played, or during the playback of the selected sound extract, the thumbnails displayed can become interactive, so that the user can interact, by means of a man-machine interface such as described above, with one or more of these thumbnails.

[0140] Alternatively, provision can be made for the displayed thumbnails to be interactive even before the selected sound extract is played, and that any interaction with a displayed thumbnail occurring before the selected sound extract is played results in the emission of a error message, or is interpreted as an indication in favor of identifying the user as being a machine.

It is considered that the user interacts, as proposed to him, with at least one thumbnail of the set of thumbnails. [0142] Accordingly, a selection signal, or several selection signals, indicating a user interaction by means of a human-machine interface with one or more thumbnails of the set of thumbnails is received (S5). For example, such a selection signal can be transmitted to the server side in order to be interpreted there. This second selection signal comprises one or more second identifiers.

[0143] Based on the received selection signal, or received selection signals, a second subset of thumbnails is determined (S51). This second subset of thumbnails is formed of the thumbnail with which the user has actually interacted, or the thumbnails with which the user has actually interacted.

At this stage, therefore, we have:

- a first subset of thumbnails, formed exclusively of the thumbnail or thumbnails with which a human user was expected to interact because these thumbnails are associated with the selected sound clip, and

- a second subset of thumbnails, consisting exclusively of the thumbnail(s) with which the user actually interacted during or after the broadcast of the selected sound clip.

[0145] This results in:

- a first subset of first identifiers of a first sound, and

- at least one second identifier received from a man-machine interface, when hearing a second sound by a user of the man-machine interface.

This first subset of first identifiers and this or these second identifiers are then compared (S6) with a view to identifying (S7) the user as being appropriate or as not being appropriate, according to predefined rules.

The comparison can return several types of results.

For example, the result of the comparison can indicate that the first and the second subset of thumbnails are identical. This means that the user has correctly identified all the thumbnails that he was asked to associate with the selected sound clip. An example of a predefined rule can provide that when the comparison returns such a result, then the user is identified as being a human.

For example, the result of the comparison may indicate that the first and the second subset of thumbnails have no element in common. This means that the user has certainly interacted with one or more thumbnails, but has not correctly identified any of the thumbnails that he was asked to associate with the selected sound clip. An example of a predefined rule can provide that when the comparison returns such a result, then the user is identified as being a machine.

In addition to these binary examples, other cases are possible. A possible example of a predefined rule is to identify the user as a machine when the first and the second subset of thumbnails are not strictly identical.

[0151] Alternatively, additional rules can be predefined to take account of certain cases where the first and the second subset of thumbnails partially overlap.

For example, the result of the comparison may indicate that the first subset of thumbnails includes all the thumbnails of the second subset of thumbnails, as well as one or more additional thumbnails. The user therefore correctly selected only part of the thumbnails which were expected to associate them with the selected sound clip, and did not select any other thumbnail.

[0153] In such a situation, a predefined rule can, for example, require direct conclusion of the human-machine differentiation by identifying the user as being a human, in order to pursue a general objective of ergonomics.

[0154] Alternatively, if the main objective is to prevent a machine from being incorrectly identified as being a human as much as possible, a predefined rule can for example provide:

- to trigger a new iteration of the algorithm by selecting a new sound extract and a new set of thumbnails, or even

- to implement, in addition, another form of Turing test to definitively identify the user as a human or as a machine.

The algorithm described above can make it possible, optionally, not only to identify a user as being a human or as being a machine, but also to enrich a database associating sound clips with thumbnails representing each a source entity of such a sound clip. Such a database can for example be used downstream to perform sound recognition from recordings.

For example, if the user is identified as being a human, then it is possible to trigger a new iteration of the algorithm by selecting a new sound clip and a new set of thumbnails.

[0157] In this new iteration, it is always necessary to ensure:

- that one or more of the selected thumbnails represent an entity that can be uniquely associated by a human with the selected sound clip, and

- that one or more of the remaining thumbnails represent an entity that can be considered by a human, unequivocally, as unrelated to the selected sound clip.

Furthermore, in this new iteration, provision is also made, among the thumbnails selected, for one or more thumbnails representing an entity whose relationship with the selected sound extract is indefinite, potentially equivocal for a human.

Consider for example a selected sound extract, from a database of sound extracts and corresponding to the noise of a mower engine.

This selected sound extract can be associated, in a correspondence table, with a first group of thumbnails from a database of thumbnails and representing mowers.

It is also possible to define, in this thumbnail database, a second group of thumbnails representing animals, musical instruments, or other entities that are not likely to emit a sound resembling the sound of a lawn mower engine. The thumbnails of this second group of thumbnails can be associated, in the correspondence table, with an indication according to which they are unrelated to the selected sound extract.

Finally, it is also possible to define, in this thumbnail database, a third group of thumbnails representing another device fitted with a motor, for example a moped, and capable as such of emitting a sound similar to that of a lawnmower engine.

The selection of the set of thumbnails can thus provide, in this new iteration, for selecting one or more thumbnails from each of the three groups of thumbnails defined above.

[0164] The set of vignettes thus comprises:

- a first subset of thumbnails formed of the thumbnails selected within the first group of thumbnails (here thumbnails representing a mower),

- another subset of thumbnails formed of the thumbnails selected within the second group of thumbnails (here thumbnails representing various entities, for example animals or musical instruments), and

- A third subset of thumbnails formed of the thumbnails selected within the third group of thumbnails (here thumbnails representing mopeds).

This third subset of vignettes is determined (S31) as designating a subset of vignettes whose relationship with the selected sound extract is undefined at the time of their selection.

[0166] The interaction with the user is the same as during the previous iteration of the algorithm.

The selected set of thumbnails is displayed, the selected sound extract is played, then at least one selection signal is received.

This signal comes from a man-machine interface and is indicative of a selection of one or more thumbnails from the set of thumbnails displayed. [0169] After identification, on the basis of this selection signal received, of a second subset formed of each selected thumbnail, it is possible to carry out a comparison (S61) between the third subset of thumbnails and the second subset of thumbnails. In this example, it is therefore determined whether or not the thumbnails selected by the user represent mopeds.

For each thumbnail (here representing a moped) whose relationship with the selected sound clip is initially uncertain, this makes it possible to determine whether the user, identified as being a human, considers that this thumbnail represents, or not, a entity capable of emitting a noise corresponding to such a sound extract.

If the result of such a comparison (S61) indicates that the second subset of thumbnails (those selected) includes all the thumbnails of the third subset of thumbnails (those representing a moped), then provision can be made for associating (S71) the selected sound extract with the thumbnails of the third subset. It is also possible, in addition, to carry out a comparison (S6) between the first subset of thumbnails (here those representing a mower) and the second subset of thumbnails. If the result of such a comparison (S6) indicates that the second thumbnail subset does not include any thumbnail of the first thumbnail subset (in this example, representing a lawn mower), then this means that the selected sound clip was incorrectly associated with the thumbnails of the first subset of thumbnails. Therefore, provision may be made, in such a case, to update a pre-existing correspondence table between sound extracts and vignettes. Such an update comes down, for example, to associating the selected sound clip no longer with the first subset of thumbnails (here those representing a mower), but instead with the third subset of thumbnails (here those representing a moped).

We now refer to Figure 4 which shows an example of a variant of the graphical interfaces of Figures 1 and 2.

In this example, a plurality of play buttons (11) are displayed, here nine play buttons distributed in the form of a 3×3 grid. When a user interacts with one of these playback buttons, a sound clip associated with it is played. Simultaneously, the play button with which the interaction is currently performed can animate, for example in the form of an emphasis.

[0175] Once the user has interacted with all the playback buttons, he can interact with a "Next" button (14) to trigger an animation of the graphical interface.

A set (2) of vignettes is then superimposed on the play buttons (11). In other words, each play button is replaced with a corresponding thumbnail.

Simultaneously, an image (13) representing an entity likely to have generated one of the sound clips having been broadcast to the user is displayed. This image (13) is associated with the sound clip in question.

[0178] The user is then asked to indicate the sound extract, among those having been broadcast, likely to have been emitted by the entity represented in the image (13). For this, it is expected that the user interacts with the thumbnail which is located in the same place as the play button that triggered the sound clip in question, then, optionally, validates his choice by interacting with a validation button (15 ).

We now refer to Figure 5 which illustrates an example of a variant of the algorithm shown in Figure 3, this variant being linked for example to the graphical interface shown in Figure 4.

The algorithm illustrated in Figure 5 incorporates all the essential elements of that illustrated in Figure 3, which are described again below, and can also incorporate any of the optional elements shown in Figure 3.

Several sound clips are selected (S1), for example from a database of sound clips.

[0182] An image associated with one of the selected sound extracts and representing an entity likely to have emitted, or capable of emitting, this sound extract selected, is also selected (S2), for example in an image database.

A subset of sound extracts is determined (S3), it is defined as designating the sound extract associated with the selected image.

[0184] Several read buttons are displayed simultaneously (S41). These buttons are interactive and respectively associated with one of the selected sound clips. A user can therefore act successively with each play button and thus listen successively to each of the selected sound extracts.

[0185] Concretely, following the interaction of the user with a play button, a signal indicating this interaction is received (S42) and triggers (S43) the broadcasting of the sound extract associated with this play button.

[0286] Once all the selected sound clips have been broadcast, it is possible to trigger an animation routine of the graphical interface, in order to display (S4) a set of thumbnails superimposed on the play buttons, as well as the selected image.

Among the thumbnails of the set of thumbnails, it is possible to define a first subset of thumbnails as formed of the thumbnail superimposed on the play button having triggered the sound clip associated with the selected image.

[0188] For example, if the selected image represents a cat and if the interaction with the play button located at the top left of a grid of 3x3 play buttons has triggered the emission of a sound extract, associated to the selected image and corresponding to a meow, then the first subset of thumbnails is formed of the thumbnail located at the top left of a grid of 3x3 thumbnails.

[0289] The user is then invited to indicate the sound extract among those broadcast likely to be emitted by the entity represented on the selected image.

To do this, the user is invited to interact, specifically, with the thumbnail having the same position as the play button that triggered the sound clip associated with the selected image. [0191] Following the user's interaction with a thumbnail, a signal indicating this interaction is received (S5) and based on this signal, a second subset of thumbnails is determined (S51). The second thumbnail subset is the thumbnail with which the user actually interacted.

Next, a comparison test can be carried out (S6) between, on the one hand, the first subset of thumbnails and, on the other hand, the second subset of thumbnails and, on the basis of the result of this test, identifying (S7) the user either as a human or as a machine.

Of course, the algorithm shown in Figure 5 can alternatively, as an additional variant, require:

- to identify several sound extracts associated with the same selected image,

WHERE

- to identify all the sound clips associated with any of the images from a list of selected images, or even

- identify one or more sound clips not associated with a selected image.

Reference is now made to FIG. 6, which represents an example of a system capable of implementing any of the algorithms described above.

The system comprises a first processing circuit (100) and a second processing circuit (200) connected through a network (300), for example the Internet network or a local network.

The first processing circuit comprises at least one processor (101) connected to a memory (102) and to a network communication interface (103)

The first processing circuit (100) can for example be integrated into a desktop or portable computer, into a network gateway, into a computer server or, more generally, into any computer or network equipment.

The second processing circuit (200) also comprises at least one processor (201) connected to a memory (202) and to at least one communication interface (203), in particular a network communication interface, thereby enabling communication with the first processing circuit (100) via the network (300).

The second processing circuit (200) designates a network equipment such as a decoder box or a terminal such as a multifunction telephone and is intended to be used to provide a service to one or more users.

[0200] The second processing circuit (200) is also connected to a visual reproduction peripheral (204) such as a screen or a video projector to display a graphic interface, a sound reproduction peripheral such as a loudspeaker which may optionally be integrated into the visual reproduction peripheral, and a man-machine interface (205), for example voice or touch, via which a user can interact with the graphical interface displayed.

We now refer to Figure 7, which illustrates yet another example of a variant of the graphical interfaces shown in Figures 1, 2 and 4.

[0202] As in Figures 1, 2, and 4, the graphical interface of Figure 7 includes a button (11) for controlling playback of a second sound, a series (2) of interactive elements during the audition of the second sound, and a reset button (31).

[0203] The series (2) of interactive elements in Figure 7 differs from that of Figures 1 and 2 in that it is devoid of thumbnails but displays a selection of words.

[0204] Thus, when the user listens to a second sound, the latter interacts with the graphical interface by means of a man-machine interface by selecting one of the words displayed, thus generating a second identifier.

[0205] The graphical interface shown in Figure 7 meets the need to determine whether or not a user has a certain type of particular equipment.

[0206] Provision can be made, for example, for the user to be considered to be an appropriate user as soon as he is equipped with a sound reproduction device making it possible to reproduce the second sound. [0207] Thus, it suffices to allow the user to:

- control the reading of the second sound, for example by means of the related control button (11), then

- characterize this second sound by an interaction with one or more interactive elements of the series (2) of interactive elements.

[0208] With this in mind, the nature of the series (2) of interactive elements does not matter since: that these elements are respectively associated with a first identifier associated with a first sound and that interacting with the one of these interactive elements generates a second identifier.

[0209] There is therefore no need for the interactive elements to be, in particular, thumbnails.

[0210] Alternatively, it can be provided for example that the user is considered to be an appropriate user when he is equipped with a sound sensor making it possible to pick up the second sound.

[0211] Thus, it suffices to allow the user, when hearing the second sound, to:

- trigger the capture of this second sound, for example by means of a control button, and

- characterize this second sound by an interaction with one or more interactive elements of the series (2) of interactive elements.

These two steps do not need to be implemented in any particular order.

The second sound can then be compared to a database of first sounds so as to identify therein a first sound having a resemblance to the second sound.

[0214] At least one second identifier, resulting from the interaction of the user to characterize this second sound, can then be compared to a first subset of first identifiers associated with the first sound thus identified, for characterize, according to a predefined rule, the user as being or not being an appropriate user.

[0215] For example, the result of this last comparison can make it possible to identify the user as being equipped with a sound sensor, and consequently to determine that the user is an appropriate user.

Claims

Claims [Claim 1] A method of characterizing a user, the method being implemented by a processing device, the method comprising: - a (S6) comparison between at least a first identifier of a first sound and at least a second identifier received from a man-machine interface, the second identifier being received after a reproduction of a plurality of second sounds by the man-machine interface, the comparison triggering, if the second identifier received corresponding to the first identifier, a characterization of the user as being an appropriate user. [Claim 2] A method according to claim 1, wherein the method comprises: - reception (S5) of at least one signal originating from a man-machine interface, the signal comprising the second identifier of the second sound heard. [Claim 3] A method according to any of claims 1 or 2, wherein the method comprises one of the following steps: - a command for reproducing a sound extract comprising a first sound associated with one of the first subsets of first identifiers of first sounds, the second sound heard comprising the first sound reproduced; - a command for capturing the second sound heard, the capture triggering a comparison of the second sound heard with at least one first sound associated with a first subset of identifiers. [Claim 4] A method according to any of claims 1 to 3, wherein the method comprises: - a command (S4) for displaying a set of thumbnails, the set of thumbnails displayed comprising at least one identifier of a first subset associated with the second sound heard, [Claim 5] A method according to any of claims 1 to 4, wherein the method comprises: - on the basis of at least one selection signal received comprising at least one second identifier received from a man-machine interface, during the hearing of a second sound by the user of the man-machine interface, a generation (S51) of a second subset comprising the second identifiers, - the (S6) comparison being carried out between at least a first subset of first identifiers of a first sound and at least a second identifier of the second subset, the comparison triggering, if at least a second identifier or all the second identifiers of the second subset received belong to the first subset associated with a first sound corresponding to the second sound heard, characterizing the user as being an appropriate user. [Claim 6] A method according to any of claims 1 to 5, wherein the first identifier and the second identifier have one of the following identifier types: - at least one word, - at least one image, - a label associated with an object; - a thumbnail. [Claim 7] A method according to any of claims 1 to 6, wherein: - the second sound being a sound produced by a given element, each first identifier of the first subset of first identifiers associated with the first sound corresponding to the second sound represents the given element. [Claim 8] Method according to any one of claims 1 to 7, comprising a selection (S1) of the second sound in a base of sound extracts on the basis of a correspondence table associating, with each sound extract of the base , a respective class. [Claim 9] Method according to any one of claims 1 to 8, comprising a selection (S2) of said at least one first subset of first identifiers in a base of first identifiers on the basis of a correspondence table associating, at each first identifier of the base, a respective class. [Claim 10] A method according to claim 9 when dependent on claim 8, wherein the selection (S2) of said at least a first subset of first identifiers in the first identifier database is further based on the class associated with the sound clip selected as the second sound. [Claim 11] A method according to any one of claims 8 to 10, the set of thumbnails further comprising a third subset formed of at least one thumbnail, the method comprising: - a comparison test (S61) between the third subset and the second subset, and - on the basis of this test, an update (S71), in the correspondence table associating, with each sound extract from the base of sound extracts, a respective class, of the class associated with the selected sound extract. [Claim 12] A method according to any one of claims 8 to 10, the set of thumbnails further comprising a third subset formed of at least one thumbnail, the method comprising: - a comparison test (S61) between the third subset and the second subset, and - on the basis of this test, an update (S71), in the correspondence table associating, with each thumbnail of the thumbnail base, a respective class, of the class associated with at least one thumbnail of the third subset . [Claim 13] Computer program comprising instructions for implementing the method according to any one of Claims 1 to 12 when this program is executed by a processor. [Claim 14] Non-transitory computer-readable recording medium (102) on which is recorded a program for carrying out the method according to any one of claims 1 to 12 when this program is executed by a processor. [Claim 15] Processing device capable of implementing user characterization processing on command from an interface man-machine, in which the processing device comprises: - a comparator able to compare at least a first identifier of a first sound and at least a second identifier received from a man-machine interface, the second identifier being received after a reproduction of a plurality of second sounds by the interface man-machine, the comparator triggering, if the second identifier received corresponding to the first identifier, a characterizer capable of characterizing the user as being an appropriate user.