DEVICE FOR CHANGING AN EMOTIONAL STATE OF A USER

Description

Technical field of the invention

[0001] The present invention relates to a device and a method for modifying an emotional state of a user. It applies, in particular, to the field of improving well-being and controlling the emotional state of an individual. State of the art

[0002] The development of cognitive, motor, and sensory skills is part of a desire to increase human life expectancy. In this quest for performance and well-being, neurotechnology is an essential solution. This field of research results from the convergence of neuroscience and computer science, which is at the origin of modifications of mental schemas and models.

[0003] One of the goals of neurotechnologies is to enhance human performance and improve well-being. Such enhancements and improvements are possibly achieved by modifying, for example, human emotional stress.

[0004] Emotional stress is a natural, non-pathological reaction of the human body to environmental stimuli (stressors), for example. Such stress is therefore an archaic natural defense mechanism that can apply to all human beings, without being associated with an illness.

[0005] Currently, the following solutions are known: Melomind (registered trademark): a system consisting of a sound headset delivering sounds in order to directly monitor an individual's relaxation level, Muse (registered trademark): a system consisting of EEG (for "electroencephalogram") sensors measuring tiny electrical fields in the brain, heart rate, respiration and oximetry sensors, Dreem-Urgo (registered trademark): a system consisting of a physiological data sensor and EEG sensors and Emotiv (registered trademark): a system consisting of an EEG sensor and earphones allowing an indication of the user's stress and distraction level. However, all these solutions have the following drawbacks: Failure to take into account user preferences in sound stimuli: The algorithms for providing sound stimuli only take into account the individual's brain activity by imposing noises to which they are not accustomed, resulting in disinterest on the part of the individual or even rejection of solutions. No real-time adaptability: even if these technologies perceive brain activity signals in real time, the algorithms used are incapable of adapting to unanticipated variations in the user's wishes throughout their use. They have been designed in such a way that the user goes from point A to point B by a single path, often based on: brain activity and / or the individual's physiology (heart rate). Highly constrained emotional trajectories: Currently, the user cannot target the emotional state they wish to achieve.Indeed, if he buys a Melomind or Muse headset, it is only with the objective of relaxing. While some systems like Emotiv allow reading different emotions, they do not allow the user to move from one emotion to another. These technologies are often unidirectional. They start from stress to arrive at a state of relaxation or concentration. The reverse is therefore not possible.

[0006] Document CN 110 947 076 is known, disclosing a smart wearable brainwave music device for regulating mental state. Document US 2019 / 060 605 is also known, which discloses a device for modifying a user's cognitive state. However, none of these devices allow for adaptation of the sound files played according to a user's choices in real time.

[0007] In addition, we know of document US 2018 / 027 347 which discloses a sound analysis system to automatically predict the effect that these sounds will have on the user. However, this device does not allow a combination between detection of the user's emotional state in real time and adaptation of the sound files played according to the user's choices in real time. Subject of the invention

[0008] The present invention aims to remedy all or part of these drawbacks.

[0009] To this end, according to a first aspect, the present invention relates to a device as defined in claim 1 and its dependent claims. It relates to a device for modifying an emotional state of a user, which comprises: a real-time reader of electroencephalographic signals, a module for determining an emotional state based on a read electroencephalographic signal, a means for determining a target emotional state, an automatic selector of a succession of at least one sound file, from a pre-constituted list of sound files, based on the determined target emotional state, the read electroencephalographic signal and at least one parameter associated with each said sound file, an electroacoustic transducer configured to read the selected succession of sound files, a secondary selector of a sound file and a means for updating the succession based on the sound file manually selected by the secondary selector.

[0010] With these provisions, a user can voluntarily determine an emotional state to be achieved, the device determining an optimal vector of sound files and playing each sound file in succession to gradually adapt the user's emotional state.

[0011] The use of a real-time reader of electroencephalographic signals allows dynamic adaptation of the succession depending on its success on the evolution of the individual's emotional state or his actions, such as, for example, the unplanned playback of a sound file requiring an update of the sound file vector.

[0012] Using a pre-built list of sound files, including those produced by artists preferred by the user, helps minimize the risk of the user rejecting the device.

[0013] Furthermore, the invention allows the user to voluntarily select a sound file to be played, interrupting the selected succession, while allowing the device to adapt the succession of sound files to this interruption.

[0014] In embodiments, the device which is the subject of the present invention comprises: a sound file identifier collector and a sound file classifier configured to associate, with at least one sound file identifier, a parameter representative of an emotional state promoted by said sound file.

[0015] These embodiments make it possible to associate with sound files from a pre-constituted list indicators used in the selection of the succession of sound files to be played.

[0016] In embodiments, the classifier is a trained machine learning system.

[0017] These embodiments allow automatic classification of new sound files, ensuring efficient updating of the pre-built sound file list.

[0018] In embodiments, the trained machine learning system is a supervised neural network configured to receive parameter values as an input layer and emotional state indicators corresponding to the input layer as an output layer.

[0019] In embodiments, the classifier is configured to classify a sound file by assigning a value to at least one of three features: valence, arousal and dominance, at least one characteristic being implemented to determine an emotional state promoted by a soundtrack.

[0020] In embodiments, the machine learning system is further pre-trained through the use of a non-user-specific dataset.

[0021] Thanks to these provisions, the machine learning system is pre-trained, for example, before using the device with external data. Additional training of the learning system is therefore carried out. Thus, the pre-training of the learning system is reinforced.

[0022] In embodiments, at least one sound file is associated with an indicator of a user behavior with respect to each said sound file, the automatic selector being configured to select a succession of at least one sound file according to a value of said indicator for at least one sound file.

[0023] These embodiments allow quantifying a user's preference for a sound file in order to minimize the risk of rejection of the selected succession.

[0024] In embodiments, the user behavior indicator is a parameter representative of a number of listens and / or a number of playback interruptions in favor of another audio track.

[0025] Thanks to these provisions, a determination of the behavior indicator is easily carried out.

[0026] In embodiments, the automatic selector comprises a sound file filter based on at least one indicator of user behavior with respect to at least one sound file, the selector being configured to select a succession of sound files from a list of sound files filtered by the filter.

[0027] These embodiments allow quantifying a user's preference for a sound file in order to minimize the risk of rejection of the selected succession.

[0028] According to the invention, a parameter used by the automatic selector to select a succession of sound files is an indicator representative of an emotional state value associated with at least one sound file.

[0029] In embodiments, a parameter used by the automatic selector to select a succession of sound files is, in addition, a technical parameter selected from the duration, mode, key, time signature and tempo of the sound file.

[0030] Thanks to these provisions, a selection made by the automatic selector depends on the technical parameters inherent to the sound file. Thus, the automation of the device is reinforced.

[0031] In embodiments of the invention, the succession of sound files is configured to exhibit an increasing gradient of emotional state value corresponding to the determined target emotional state.

[0032] These embodiments make it possible to determine a succession of increasing intensity, associated with a target emotional state.

[0033] In embodiments, the real-time reader of electroencephalographic signals is non-invasive.

[0034] In embodiments, the reader is an electroencephalogram type headset.

[0035] These provisions make the device easier to use. Furthermore, the user's bodily integrity is maintained when using the device. In other words, there is no physical discomfort when using the device. Furthermore, when the player is a headset, the device is mobile according to the user's movements.

[0036] According to a second unclaimed aspect, the present description relates to a method of modifying an emotional state of a user, which comprises: a step of determining a target emotional state, then, iteratively, at least some of the following steps: a step of reading electroencephalographic signals in real time, a step of determining an emotional state as a function of a read electroencephalographic signal, a step of automatically selecting a succession of at least one sound file, from a pre-constituted list of sound files, as a function of the determined target emotional state, the read electroencephalographic signal and at least one parameter associated with each said sound file, a step of reading, by an electroacoustic transducer, the selected succession of sound files, a step of secondary selection of a sound file and a step of updating the succession as a function of the sound file manually selected by the secondary selector.

[0037] The advantages, aims and particular characteristics of the method which is the subject of the present invention being identical to those of the device which is the subject of the present invention, they are not recalled here. Brief description of the figures

[0038] Other advantages, aims and particular characteristics of the invention will emerge from the following non-limiting description of at least one particular embodiment of the device and method which are the subject of the present invention, with reference to the appended drawings, in which: there figure 1 represents, schematically, a first particular embodiment of the device which is the subject of the present invention, the figure 2 represents, schematically, a second particular embodiment of the device which is the subject of the present invention, the figure 3 represents, schematically and in the form of a flowchart, a particular succession of steps of the method which is the subject of the present invention, the figure 4 schematically represents a distribution of a sample of music files in terms of energizing nature, distributed between values of 0.1 and 1, the Figure 5schematically represents a distribution of a sample of music files in terms of dance nature, distributed between values of 0, 1 and 1, the figure 6 schematically represents a distribution of sound files to be read corresponding to a determined logarithmic vector of modification of emotional state, the figure 7 schematically represents a distribution of sound files to be read corresponding to a determined linear vector of modification of emotional state, the figure 8 represents, schematically, a distribution of acoustic values of sound files corresponding to distinct emotional states and the figure 9 schematically represents a distribution of volume values of sound files corresponding to distinct emotional states. Description of the embodiments

[0039] This description is given without limitation, each characteristic of an embodiment being able to be combined with any other characteristic of any other embodiment in an advantageous manner.

[0040] Please note that the figures are not to scale.

[0041] An "emotional state" is the result of the interaction of subjective and objective factors, carried out by neural or endocrine systems, which can: induce experiences such as feelings of arousal, pleasure, or displeasure; generate cognitive processes such as perceptually relevant reorientations, appraisals, and labeling; activate global physiological adjustments; induce behaviors that are, most often, expressive, goal-directed, and adaptive.

[0042] Such an emotional state is, for example, for a human: Stress (difficulty concentrating, chaotic and disrupted thought processes due to the source of stress, excitement, fear), Relaxation (calm state, lack of or low thinking activity, ideas flow slowly), Concentration (intense brain activity, non-receptiveness to external signals) and Interest or engagement (measures the level of immersion in any stimulus. Engagement can increase when immersed in stimuli that can be positive or negative in nature. It has been found in experimental brain research that engagement decreases in boring, mundane and automatic cognitive processes).

[0043] We observe, on the figure 1 , which is not to scale, a schematic view of an embodiment of the device 100 which is the subject of the present invention. This device 100 for modifying an emotional state of a user comprises: a real-time reader 105 of electroencephalographic signals, a module 110 for determining an emotional state as a function of a read electroencephalographic signal, a means 115 for determining a target emotional state, an automatic selector 120 of a succession of at least one sound file, from a pre-constituted list of sound files, as a function of the determined target emotional state, the read electroencephalographic signal and at least one parameter associated with each said sound file and an electroacoustic transducer 125 configured to read the selected succession of sound files.

[0044] The reader 105 is, for example, an electroencephalogram type headset equipped with earphones acting as an electroacoustic transducer 125. The type of electroencephalogram considered may be of any type known to a person skilled in the art of neuro-technologies. Preferably, a non-invasive type electroencephalogram is implemented.

[0045] An electroencephalogram has the function of capturing electrical signals resulting from the summation of synchronous postsynaptic potentials from a large number of neurons. Such a signal can be representative of a neurophysiological activity of the individual wearing the reader 105 and, thus, of an emotional state of this individual.

[0046] The acquisition is for example made from seven dry electrodes placed on the skull of a user, preferably in positions A1, T3, C3, CZ, C4, T4, A2, T5 and T6 according to the 10-20 system. These electrodes measure the potential difference (in Volts) between the different locations and "the ground", placed on the nearest ear.

[0047] The choice of positioning these electrodes is mainly linked to the geometry of the helmet and the comfort of use, but also to the selection of certain points less subject to motor artifacts (eyelid blinks, sniffs, etc.). The electrodes are connected to a digital card. This card is configured to transmit the signal to the determination module 110 or to a computer, for example via Bluetooth to a USB receiver connected to the computer.

[0048] The chosen sampling duration is for example 5 seconds (every 5 seconds, we retrieve the signal of the last 5 seconds passed). This value could be increased to 8 seconds, or even 10 seconds, 5, 8 and 10 being the sample durations allowing the best inference of emotions.

[0049] From the raw signal, the intensity of the different frequencies can be calculated by Fourier transform. A first pre-processing of the signal can be applied to remove frequencies close to 50 Hz (49 to 51 Hz) or 60 Hz (59 to 61 Hz) which are intensely interfered with in the presence of electrical devices connected to the mains near the recording device (headphones).

[0050] A second bandpass filter can be applied to keep only frequencies in the 2 to 58 Hz range, in order to eliminate low frequency noise, and high frequency gamma bands (60 Hz and above) that Bluetooth does not allow us to describe properly.

[0051] The two filters used are, for example, of the Butterworth type of order 5. The determination module 110 is, for example, a computer software program executed by an electronic calculation circuit. This determination module 110 is configured, for example, to determine an emotional state of the individual in the following manner: The model for describing emotions is chosen to be the three-variable system commonly used called the "VAD" model for the axes of Valence, Arousal (translated as "excitation"), and Dominance in English.

[0052] Valence describes the negative, neutral, or positive nature associated with an emotion.

[0053] Arousal measures the passive, neutral, or active nature of the emotional state described.

[0054] Dominance describes the assertive or submissive nature of the emotional state described. This axis allows, for example, to discriminate between rage and fear (both of which are characterized by low valence and high arousal), or relaxation and joy.

[0055] For each axis, we give ourselves 3 possible discrete values: Valence: -1 for negative, 0 for neutral, 1 for positive, Arousal: -1 for passive, 0 for neutral, 1 for active and Dominance: -1 for low, 0 for medium, 1 for high.

[0056] The following labels are assigned to the VAD value triplets: Excited: (1, 1, 1) Happy: (1, 0, 1) Content: (1, 0, 0) Relaxed, Unwinded: (1, -1, 0) Calm: (0, -1, 0) Sad, Depressed: (-1, -1, -1) Stress-free: (0, 1, 0) Neutral: (0, 0, 0) Deeply sad: (-1, 0, -1)

[0057] Other emotional descriptors can be used in addition to VAD values, as they are easy to detect and recognize from brain recordings, such as relaxation and concentration.

[0058] Each sound track can be placed according to these coordinates so as to calculate a vector from one place to another allowing, starting from determined coordinates, to wait for other coordinates corresponding to a target emotional state.

[0059] The determination module 110 can be implemented locally or remotely and accessible via a data network. For example, the determination module 110 can be implemented on a smartphone connected to the reader 105 wired or, preferably, wirelessly.

[0060] The means 115 for determining a target emotional state is, for example, a human-machine interface (for example of the graphical interface type associated with an input device) or a software interface (for example of the API type, for Application Programming Interface). This determination means 115 is configured to receive, as input, a variable signal between several possible emotional states. These emotional states can be predetermined, that is to say form a finite list of possibilities from which an implemented interface selects. These emotional states can be determined according to the content of an input made on an interface.For example, a human-machine interface allows the free entry of alphanumeric characters representative of a human language, a user of the device 100 entering keywords describing an emotional state to be achieved, determination means 115 being configured to associate defined emotional states with these keywords.

[0061] The determination means 115 may be implemented locally or remotely and accessible via a data network. For example, the determination means 115 may be implemented on a smartphone associated with the determination module 110.

[0062] The automatic selector 120 is, for example, computer software executed by an electronic calculation circuit. The automatic selector 120 is configured, for example, to execute an algorithm measuring a distance between the read emotional state of a user of the device 100 and the target state determined by the determination means 115. Depending on the distance thus measured, a succession of at least one sound file is selected according to at least one parameter associated with at least one said sound file.

[0063] Such a parameter can be, for example: a technical parameter, such as the duration, mode, key, time signature, or tempo of the sound file, or an acoustic or psychoacoustic parameter representative of: the acoustic nature of the sound file, i.e., whether or not electronic instruments are used and / or the proportion of such use, the energizing, or invigorating, nature of the sound file, i.e., a perceptual measure of loudness and activity - perceptual features contributing to this attribute include dynamic range, perceived loudness, timbre, rate of occurrence, and overall entropy, the instrumental nature of the sound file, i.e., whether or not voices are used on the sound file, the danceable nature of the sound file, measured, for example, by tempo, rhythm stability, beat strength, and overall regularity of the audio file, the valence of the sound file,i.e. the positivity of the sound file, the nature of the recording of the sound file, i.e. studio recording or during a live performance of the sound file, the nature of the speech density, i.e. the proportion of words and music in a sound file and / or the intensity of the sound file, i.e. the average intensity, measured in decibels, of the sound file.

[0064] Each of these parameters can be directly associated with an emotional state and thus, depending on the target emotional state, be a candidate for inclusion in the succession of sound files.

[0065] The association between values for these parameters and emotional states (via, for example, a VAD profile) can be achieved, for example, by implementing a learning algorithm, obtained in a similar way to the IADS-E dataset from the Center for the Study of Emotions and Attention at the University of Florida.

[0066] Alternatively, an expert system can be implemented, associating particular values of the VAD profile with an emotional state. An example of such an implementation is provided above.

[0067] In a simplified way, energizing nature and dancing nature are associated with excitement, mode and valence with valence and intensity with dominance.

[0068] For each sample, we then construct, for example, a statistical model of each acoustic descriptor. The chosen model is, for example, a Gaussian mixture, i.e. a weighted set of one to five Gaussian curves ("bell curves") whose means and standard deviations are recorded, and the weights associated with each Gaussian as well. The resulting mixed Gaussian model describes a probability density curve, which associates with each value of the acoustic parameter considered the probability of being observed for an audio track of the given group (high or low valence, high or low arousal, high or low dominance).

[0069] We thus obtained an approximation of the probability that an audio track with given acoustic characteristics is in each quadrant of the VAD space.

[0070] The average probability of belonging to the positive and negative quadrant is calculated for each axis and for each audio track. This gives the coordinates of the audio file in question in VAD space. This position in VAD space will be read when determining which audio tracks to add to a playlist.

[0071] We observe, in figure 4 , schematically, a distribution of a sample of music files in terms of energizing nature, distributed between values of 0.1 and 1.

[0072] We observe, in Figure 5 , schematically, a distribution of a sample of music files in terms of dance nature, distributed between values of 0, 1 and 1.

[0073] We observe, in figure 8 , schematically, a distribution of the quantification of the acoustic nature of a sample of sound files corresponding to, in particular, two distinct states: a calm emotional state 805 and an emotional state of concentration 810.

[0074] We observe, in figure 9 , schematically, a distribution of the quantification of the sound volume of a sample of sound files corresponding to, in particular, two distinct states: an emotional state of concentration 905 and an emotional state of intense physical activity 910.

[0075] Preferably, a parameter used by the automatic selector 120 to select a succession of sound files is an indicator representative of an emotional state value associated with at least one sound file. Each sound file is then associated with a vector for quantifying the impact for at least one emotional state. For example, a sound file may have a first value corresponding to the impact of this sound file on a stress level of a listener and a second value corresponding to the impact of this sound file on a relaxation level of a listener.

[0076] Preferably, the succession of sound files is configured to present an increasing gradient of emotional state value corresponding to the determined target emotional state.

[0077] In other words, the current emotional states and determined targets are described by coordinates on axes that constitute all or part of the parameters listed above, in a multi-dimensional space. Such a vector can be constructed from at least one of the parameters described above.

[0078] Such a vector, in a given dimensional space, can correspond to an affine or logarithmic function.

[0079] According to a first algorithm, a theoretical straight-line trajectory between the two points in VAD space is first calculated. It does not yet physically correspond to a list of sound files. Then, the algorithm samples regularly spaced points along this theoretical line (depending on the number of desired sound files, itself a function of the duration of the desired playlist). Finally, the sound files from the database whose coordinates in this space are closest to each of the theoretical points are selected, resulting in an ordered list of sound files.

[0080] According to a second algorithm, the selection of audio tracks is done iteratively, by systematically searching for the file closest to the midpoint between the limits of the search interval. We start by calculating the theoretical midpoint between the two points (current and targeted states). We then determine the sound file in the database closest to this point in the VAD space, using its coordinates. This file makes it possible to cut the interval in two and produce two new intervals, on which we repeat the procedure, until, optionally, a maximum number of audio tracks is obtained.

[0081] The final trajectory obtained is less linear than with algorithm 1, but allows for smoother transitions between tracks.

[0082] The automatic selector 120 can be implemented locally or remotely and accessible via a data network. For example, the automatic selector 120 can be implemented on a smartphone associated with the determination means 115 and the determination module 110.

[0083] The succession of sound files is sent to an electroacoustic transducer 125. This sending can be done via the motherboard or a sound card of a smartphone interfaced with the automatic selector 120.

[0084] In particular embodiments, such as that shown in the figure 1 , the device 100 which is the subject of the present invention comprises: a collector 140 of sound file identifiers and a sound file classifier 145 configured to associate, with at least one sound file identifier, a parameter representative of an emotional state favored by said sound file.

[0085] The identifier collector 140 is, for example, computer software executed by an electronic computing circuit. This collector 140 is, for example, configured to collect the identifiers of sound files whose playback is controlled by at least one user associated with the device 100 from a third-party sound file playback application. In variants, the identifier collector 140 is software for reading metadata of sound file identifiers stored in a local or remote computer storage.

[0086] The identifier collector 140 can be implemented locally or remotely and accessible via a data network.

[0087] The classifier 145 is, for example, computer software executed by an electronic computing circuit. This classifier 145 is configured to, from parameters of the sound files, as described above, assign a quantitative value of the impact of the sound file on the emotional state of a listener of the sound file.

[0088] In particular embodiments, the classifier 145 is a trained machine learning system. Such a classifier 145 may be, for example, a supervised or unsupervised machine learning algorithm and of the deep learning type or not.

[0089] For example, such a machine learning system is a supervised neural network device configured to receive as an input layer parameter values, as mentioned above, and as an output layer emotional state indicators corresponding to the input layer.

[0090] In embodiments, the classifier 145 is configured to classify a sound file by assigning a value to at least one of three characteristics: valence, arousal and dominance, at least one characteristic being implemented to determine an emotional state promoted by a soundtrack. Optionally, concentration and / or relaxation may be assigned a value, via the classifier 145 or another attribution mechanism.

[0091] In an exemplary implementation of such a classifier 145, a computer program allows the user to report when they feel deeply, in their body, that they are in one of the emotional states listed above. If they confirm their state, the recorded sample is sent to the classification model which reinforces its learning.

[0092] Frequent use of this reinforcement tool is necessary for the model to learn properly. Before the user has used this tool sufficiently (several dozen times, with all emotions represented), the model's performance goes from random, to bad, then mediocre, and finally acceptable. It is possible to implement a pre-trained 145 classifier using a non-user-specific dataset to start with decent performance.

[0093] The model is trained to recognize an emotion not from the raw signal, but from transformations of it, which we call characteristics which are calculated from a sample recorded for a time t, on the 7 channels listed above: For each channel: spatial characteristics: the intensity of the alpha [7-13 Hz], beta [14-30 Hz] and gamma [31-90 Hz] frequency bands obtained by Fourier transform, and the differential entropy and temporal characteristics: the approximate entropy, the sample entropy, and the fractal dimension.

[0094] Taking into account all channels: Multi-Scale-Entropy features and Renyi entropy (or more precisely, a non-parametric estimate of Renyi entropy).

[0095] A classification algorithm can be an ensemble method (a method that averages the predictions of several classifiers) called "Random Forest", the classifiers used being decision trees. A population of one hundred decision trees is used, for example. A decision tree is a series of rules that use thresholds on the values of the features.

[0096] The algorithm's training phase involves varying these thresholds until a sufficient prediction quality is obtained. Each new sample obtained, whose associated emotional state is known, allows the thresholds to be further refined. Model performance varies from one individual to another and cannot be generally estimated. The scientific literature reports averages ranging between 60% and 85% of correct predictions, depending on the individual.

[0097] Thus, the computer program which constitutes the training means carries out the following steps: receive the raw signal from the EEG headset (5-second sample), calculate the signal characteristics, estimate the valence, arousal, and dominance values of the user's current emotional state, using an average of the decision tree predictions. transform the obtained coordinates (V, A, D) into a percentage of each emotion whose label is known, by calculating the Euclidean distance to the position of these known emotions in the VAD space and normalizing the inverse of these distances, if the user signals, via an input device (keyboard, touchscreen), that he is in a known emotional state, offer him to use the last sample to participate in the training of the model and if he validates, adjust the decision tree thresholds using this new sample and wait for the reception of a new sample.

[0098] When the training program stops, the population of decision trees is saved, to be loaded at the next startup.

[0099] In particular embodiments, at least one sound file is associated with an indicator of user behavior with respect to each said sound file, the automatic selector 120 being configured to select a succession of at least one sound file based on a value of said indicator for at least one sound file.

[0100] Such a behavior indicator is, for example, a parameter representative of a number of listens, a number of playback interruptions in favor of another sound track or any other parameter representative of a user's desire. This behavior indicator can be implemented in the selection of the succession of sound files, for example by assigning a lower weighting to candidate sound files whose number of playback interruptions is higher than the average, reflecting user dissatisfaction when playing said files.

[0101] In embodiments, at least one parameter used by the automatic selector 120 is a parameter representative of a musical proximity between sound files. Such musical proximity can be established, for example, based on metadata representative of a musical genre or based on one of the parameters described above, with respect to the automatic selector 120.

[0102] A musical proximity is determined based on the Euclidean distances in the parameter space (normalized by their units) exemplified above.

[0103] In particular embodiments, such as that shown in the figure 1, the automatic selector 120 comprises a sound file filter 121 based on at least one indicator of user behavior with respect to at least one sound file, the selector being configured to select a succession of sound files from a list of sound files filtered by the filter.

[0104] Such a filter 121 is, for example, a software filter making it possible to establish a sample of sound files that are candidates for selection in the succession of sound files upstream of the actual selection.

[0105] In particular embodiments, such as that shown in the figure 1, the device 100 which is the subject of the present invention further comprises a secondary selector 130 of a sound file and a means 135 for updating the succession according to the sound file manually selected by the secondary selector. The secondary selector 130 is, for example, a human-machine interface (for example of the graphical interface type associated with an input device) or a software interface (for example of the API type). This secondary selector 130 is configured to receive, as input, an identifier of a sound file to be played. A variant of such a secondary selector 130 is, for example, a touch screen associated with a graphical user interface (GUI) allowing the entry of a sound file identifier.

[0106] This selector 130 allows a user to force the playback of a sound file, regardless of the beneficial or negative effect of this sound file on the determined target emotional state.

[0107] However, the quantification of this beneficial or negative effect allows the determination means 115 to determine a new succession of sound files allowing the achievement of the subsequent target emotional state depending on the deviation caused by the playback of the sound file selected by the secondary selector 130.

[0108] The secondary selector 130 may be implemented locally or remotely and accessible via a data network.

[0109] We observe, on the figure 2 , schematically, a particular embodiment of a technical ecosystem 200 allowing the implementation of the device 100 which is the subject of the present invention. This technical ecosystem 200 comprises: a central terminal 205, of the electronic calculation circuit type configured to execute sequences of algorithmic instructions in software form, such as for example a computer, a server or a smartphone, a database 215 of audio files accessible from the central terminal 205, an EEG headset 210 equipped with earphones allowing the reproduction of audio files, connected to the central terminal 205 so as to: receive an audio file to be read (or a location of a resource to be read on a data network) and send to said terminal 205 an indicator representative of a read EEG signal, said terminal 205 locally determining an emotional state of the wearer of the headset 210 or transmitting this indicator for remote determination, via the use of a computer server for example, a database 220 for classifying emotional states associated with sound files,associated with a computing resource (not referenced) configured to execute a classification algorithm, this database 220 being able to be fed as audio files are read by the central terminal 205 or directly connected to the audio file database 215, an interface 225 for accessing the classification database 220, by the central terminal 205 (or another computing resource), configured to produce, as a function of a determined target emotional state, a list of candidate sound files for inclusion in the succession to be produced to modify the emotional state of the user, or alternatively directly the succession of sound files to be read and optionally, a monitoring and control interface 230 allowing the display of statistics and a configuration interface of the device 100.

[0110] We observe, on the figure 3, schematically, a particular embodiment of the method 300 which is the subject of the present invention. This method 300 for modifying an emotional state of a user, comprises: a step 305 of determining a target emotional state, then, iteratively, at least some of the following steps: a step 310 of real-time reading of electroencephalographic signals, a step 315 of determining an emotional state as a function of a read electroencephalographic signal, a step 320 of automatic selection of a succession of at least one sound file, from a pre-constituted list of sound files, as a function of the determined target emotional state, the read electroencephalographic signal and at least one parameter associated with each said sound file and a step 325 of reading, by an electroacoustic transducer, the selected succession of sound files.

[0111] In embodiments, such as that shown in figure 3 , the method 300 further comprises: a step 330 of secondary selection of a sound file and a step 335 of updating the succession according to the sound file manually selected by the secondary selector.

[0112] Examples of implementation of the steps of the method 300 are described with regard to the corresponding means, as described with regard to the figures 1 And 2 .

[0113] Preferably, the means of the device 100 and of the technical ecosystem 200 are configured to implement the steps of the method 300 and their embodiments as set out above and the method 300 as well as its different embodiments can be implemented by the means of the device 100 and / or of the technical ecosystem 200.

Claims

1. Device (100) for modifying an emotional state of a user, which comprises: - a real-time reader (105) of electroencephalographic signals, - a module (110) for determining an emotional state according to an electroencephalographic signal read, - a means (115) for determining a target emotional state, - an automatic selector (120) for selecting a succession of at least one sound file, from a pre-compiled list of sound files, according to the target emotional state determined, the electroencephalographic signal read, and at least one parameter associated with each said sound file, a parameter used by the automatic selector (120) to select a succession of sound files being an indicator representative of an emotional state value associated to at least one sound file, the succession of sound files being configured to have an increasing emotional state value gradient corresponding to the target emotional state determined, and - an electroacoustic transducer (125) configured to play the selected succession of sound files, characterised in that the device also comprises: - a secondary selector (130) of a sound file, and - a means (135) for updating the succession by determining a new succession following manual selection, according to the sound file manually selected by the secondary selector to achieve the target emotional state, the new succession of sound files being configured to have an increasing emotional state value gradient corresponding to the target emotional state determined based on the indicator representative of an emotional state value associated to the sound file selected.

2. Device (100) according to claim 1, which comprises: - a collector (140) of sound file identifiers, and - a sound file classifier (145) configured to associate, to at least one sound file identifier, a parameter representative of an emotional state promoted by said sound file, at least one parameter being selected from: - a technical parameter, such as the duration, mode, tonality, quantification of the beat and tempo of the sound file, or - an acoustic or psychoacoustic parameter representative of: - the acoustic nature of the sound file, i.e. whether electronic instruments are used or not, and / or the proportion of such a use, - the energising, or stimulating, nature of the sound file, i.e. a perceptual measurement of the intensity and activity - the perceptual characteristics contributing to this attribute comprise the dynamic range, the perceived sound intensity, the timbre, the repetition rate and the general entropy, - the instrumental nature of the sound file, i.e. whether or not this sound file includes voice, - the danceable nature of the sound file, measured for example as a function of the tempo, rhythm stability, beat strength and general regularity of the audio file, - the valence of the sound file, i.e. the positivity of the sound file, - the recording nature of the sound file, i.e. whether the sound file contains a studio recording or a live performance recording, - the word density nature, i.e. the proportion of words and music in a sound file, and / or - the intensity of the sound file, i.e. the average intensity, measured in decibels, of the sound file.

3. Device (100) according to claim 2, wherein the classifier (145) is a trained machine learning system.

4. Device (100) according to claim 3, wherein the trained machine learning system is a supervised neural network configured to receive, as input layer, parameter values and, as output layer, emotional state indicators corresponding to the input layer.

5. Device (100) according to one of claims 3 or 4, wherein the classifier (145) is configured to classify a sound file by assigning a value to at least one of three characteristics: - valence, - energisation, and - dominance, at least one characteristic being utilised to determine an emotional state promoted by a sound track.

6. Device according to one of claims 3 to 5, wherein the machine learning system is also pre-trained via use of a non-specific dataset of the user.

7. Device (100) according to one of claims 1 to 6, wherein at least one sound file is associated with an indicator of a behaviour of the user with respect to each said sound file, the automatic selector (120) being configured to select a succession of at least one sound file based on a value of said indicator for at least one sound file.

8. Device (100) according to claim 7, wherein the indicator of the user's behaviour is a parameter representative of a number of listens and / or a number of plays interrupted to jump to another sound track.

9. Device (100) according to one of claims 7 or 8, wherein the automatic selector (120) comprises a sound file filter (121) based on at least one indicator of a behaviour of the user with respect to at least one sound file, the selector being configured to select a succession of sound files from a list of sound files filtered by the filter.

10. Device (100) according to one of claims 1 to 9, wherein a parameter used by the automatic selector (120) for selecting a succession of sound files is also a technical parameter selected from the duration, mode, tonality, quantification of the beat and tempo of the sound file.

11. Device (100) according to one of claims 1 to 10, wherein the real-time reader (105) of electroencephalographic signals is non-invasive.

12. Device (100) according to claim 11, wherein the reader (105) is an electroencephalogram headset.