The present invention concerns a method for the personalized enjoyment of digital transmissions, and relevant system.

More in detail, the present invention concerns a method that allows the creation and the continuous updating of personalized lists of preferred channels for different users within the family unit. The invention further concerns the system that implements the method according to the invention.

In the following, reference will be made to the only television case; however the concepts explained have to be understood as valid for any type of digital transmission.

In the world of analogical TV, the number of channels that can be received is very limited and the typical use of the remote control is limited to a few buttons (on/off, 0-9, channel forwards, channel backwards, previous channel). Vice versa, the world of the digital TV is characterised by a high number of composite services (schedule collected under a specific mark, which one usually refers to with the name of channels) accessible on the basis of a LCN ("Logical Channel Number", or any other numbering, for example of the reference territorial entity) composed by a certain number of digits (for example 0-999). Therefore, whoever is accustomed with the old analogical television must change with the digital one his habits and follow the dictations of the new technologies.

Patent document "Methods And Systems For Selecting And

Presenting Content Based On Learned User Preferences" (WO 2007/103938) teaches to learn the preferences of the user starting from the enjoyment times of the digital channels by means of a digital transmissions receiver to the end of improving its interaction with different contents providing systems. The invention provides for the acquisition and analysis both of the data relevant to the enjoyment times of the digital channels and geographical location data and user context data, so that it is possible to identify useful relationships on the basis of the comparison of the behaviors of multiple users. The identification of the preferences of the user is carried out on the basis of the observation of the periodicity of the behaviors of different users. As a consequence, the family unit utilizing the digital receiver is considered a san only user whose behavior periodicities are to be observed. Moreover, this known technique considers spaces of data relevant to the use of multiple services, besides the enjoyment of digital channels, and can identify in an explicit way the user for each of such services.

The present invention has the objective to make accessible, to any member of the family unit, its own preferred channels, among thousands of channels at disposal, avoiding him to have to remember a high quantity of several-digits numbers (to which the channels are associated through the LCN) and therefore by the pressure of the minimal amount of buttons, hopefully only one in most cases, exactly as it occurred in the analogical case.

It is object of the present invention to provide a method for the automated creation of personalized lists of digital transmissions, which solves the problems and overcomes the drawbacks of the prior art.

It is further specific object of the present invention a method for the personalized enjoyment of digital transmissions by the automated creation of personalized lists of predefined transmissions structures (usually called "channels" in the case of TV and/or radio) which solves the problems and overcomes the drawbacks of the prior art.

It is further specific object of the present invention to provide means for the implementation of the method that is object of the invention.

It is subject-matter of the present invention a method for the automated creation of personalized lists of digital transmissions, comprising the use of a digital decoder, associated to an audio and/or video player, and a control of the decoder and/or the player by one or more users, the method comprising the calculation of the enjoyment times of the digital transmissions and one or more sums of said enjoyment times according to one or more aggregation criteria, the aggregation criterium being relevant to a specific feature of said digital transmissions, the method being characterised in that it uses a memory area in communication with said decoder, and in that it executes the following steps:

A. Acquiring in said memory area one or more reference vectors sets, each reference vector being associated to an only class within a plurality of reference users classes, and each of said one or more reference vectors sets being associated to respective one or more predefined aggregation criteria, the components of each reference vector having as index a designation number within a predefined transmissions structure, for example a structure ordered by channel and/or genre, and as value a reference enjoyment time information for the relevant transmissions;

The acquisition can be made by the decoder by the data obtained from the digital transmission (e.g. television transmission) or from external device in communication with said memory area, or in any other useful way to the end of the acquisition itself. By the way, the decoder can also be realized by software and the control can be the remote control of the (e.g. TV) player. By "enjoyment" it is meant the watching and/or listening of the transmissions. The information of enjoyment time-span can be the time- span itself, the starting point (and possibly the ending point) and the duration or a percentage of a predefined time window. The sum of a set of enjoyment time-spans is defined as "aggregated enjoyment time-span".

By "aggregation criterium" any feature of the above-mentioned information is to be understood, on the basis of which it is possible to sum up the enjoyment time-spans of the same information. Examples of aggregation criterium are the channel on which a transmission is transmitted or the genre of the transmission. The aggregation criteria are relevant to the features of the digital transmission or the decoder. B. Acquiring, by said decoder, in said memory area, the commands that said control sends to the decoder as a consequence of the use by a current user among said one or more users;

C. Allocating in said memory area at least an enjoyment time-span vector, whose row indexes are associated to said designation numbers of said predefined transmissions structure and whose components represent the different accesses along time to the transmissions corresponding to said designation numbers and are given by a list of pairs each formed by an enjoyment start timestamp and an enjoyment duration of the relevant transmission, said enjoyment start timestamp and said enjoyment duration being calculated on the basis of the commands acquired in step B, said at least an enjoyment time-span vector being continuously updated on the basis of the data acquired by the decoder according to step B;

D. Allocating in said memory area one or more aggregated enjoyment time-span vectors, which are aggregated respectively according to said one or more predefined aggregation criteria, for example channels and/or genres criteria, each aggregated enjoyment time- span vector being calculated starting from said at least an enjoyment time-span vector of step C and having as values for each row the sum of the values of suitable functions of the components of said at least an enjoyment time-span vector, said one or more aggregated enjoyment time-span vectors being continuously updated following the update of step C;

In the case of a structure by channel or genres, the function acts on the corresponding rows of the enjoyment time-span vectors ordered in a respective way. Instead, if the aggregation starts from a structure and is aimed at producing aggregations on another structure, the rows whereon the function acts can be not corresponding in the vectors.

E. For one or more updates of step D, decomposing at least a vector v of said one or more aggregated enjoyment time-span vectors into at least a respective combination of Y vectors, with Y positive integer, among the vectors of the reference vectors set corresponding to the aggregation criteria relevant to vector v and including additionally the vectors of a pre-defined fictitious reference vectors set to the end of forming altogether a base for said combination, the number Y and the weights of the combination being obtained by means of an algorithm of search of optimal or sub-optimal combination among the vectors of the reference vectors set;

F. On the basis of the commands of step B and said at least a combination of Y vectors of step E, assigning at least one of said Y vectors and therefore at least a respective membership class to said current user according to a predefined assignation criterion; the reference vector relevant to said at least a respective membership class of step F providing by definition at least a corresponding pre-defined structure of transmissions associated to said at least a user membership class, and said at least a corresponding pre-defined transmissions structure is taken as user's personalized transmission structure. The set of the fictitious reference vectors can be also an empty set if there is already a base in the set of reference vectors.

Preferably according to the invention, the method executes the following further step subsequent to step F:

G. Allocating in said memory area one or more ordered vectors of preferred transmissions respectively for said one or more predefined aggregation criteria of step A, whose components are ordered on the basis of the larger enjoyment time of the transmission by the current user, according to the data acquired in step B and said at least a respective membership class of step F. The method moreover allows the users to provide to third parties, by any means (for example: memory support, digital transmission), and at any time, one or more the various type of vectors and/or data structures calculated or utilized by the method according to the invention.

Such information, for example the information associated to the watching time-span of the various transmissions according to the various aggregation criteria, have an enormous added value for companies such as, for example in the TV domain, broadcasters, ratings detection companies, advertising companies etc, because they allow for example to understand actually the ratings of the transmissions, to be able to personalize the advertising promotions as a function of the aggregated preferences of various users and/or even of the single user (for example, for a digital television provided with a Internet connection, by proposing, overlaid to the current transmission, a targeted advertisement). The users that wish to provide such information to third parties would obviously have an interest to do so only in exchange of a suitable reward (for example: compensation, discounts, etc), modes and details of the possible agreements between users and third parties being beyond the aims of the present patent specification.

Preferably according to the invention, the algorithm of step E utilizes the minimization a merit function of the value relevant to a function of distance between said a vector v and all the vectors of said reference vectors set, for example the geometrical distance or the sum of the geometrical distances or the maximum distance between said geometrical distances.

Preferably according to the invention, in correspondence of each new switching-on of the decoder, till the conclusion of step F, the current user is temporarily identified as the one associated to a default class among said reference users classes, or to the class with highest overall enjoyment time-span, or to the class that, on the basis of the data acquired in step B, has a greater probability to be the one that can be associated to said current user.

Preferably according to the invention, step A is executed at the time of the production, and or switching-on of the decoder, and/or upon specific request of said current user, the decoder receiving through the transmission signal, or possible Internet connection, or other possible systems, possible updates of said plurality of reference vectors.

It is further specific subject-matter of the present invention a method for the personalized enjoyment of digital transmissions, comprising the use of a digital decoder, associated to a player of digital transmissions, and a control of the decoder and/or the player by one or more users, as well as the use of a memory area in communication with said decoder, characterised in that it executes the steps of the method for the automated creation of personalized lists of digital transmissions according to the invention, said decoder being provided with means set up for the enjoyment of the transmissions on the basis of the reference vector associated to the current user in step F and/or the ordered vectors of preferred transmissions according to step G of the invention.

Preferably according to the invention, the method associates to virtual channels, i.e. channels to which a number is associated that is not included in the predefined transmissions structure according to one or more aggregation criteria, transmissions of other channels of the predefined transmission structure, sorting them chronologically in such a way to realize a schedule according to preferences of the current user, the system optionally subdividing the virtual channels into "thematic" and "general" channels, the virtual channels being "thematic" if the transmissions which come in succession share a congruence with respect to an aggregation criterion.

Preferably according to the invention, the method executes a channel surfing, i.e. an automated and personalized surfing of the transmissions, which proposes for the enjoyment to the current user each transmission for a predefined time-span L, according to one of the aggregation criteria defined in step A on the basis of said one or more ordered vectors of preferred transmissions of step G and/or on the basis of the class of the current user as identified in step F, and of suitable information of time correlation that can be calculated starting from said at least a enjoyment time-span vector of step C, such as for example day of the week and timeslot, and possibly also as a function of one or more among said aggregation criteria that are preferred by the current user because more frequently enjoyed and/or associated to his class, the personalized channel surfing being suitable to be interrupted on the basis of a command from said control.

Preferably according to the invention, said enjoyment occurs on the basis of one among said one or more predefined aggregation criteria, chosen on the basis of the data collected in step B.

It is specific subject-matter of the present invention a computer program, characterised in that it comprises code means set up to execute, when they operate on a computer, in particular a computer comprised in a decoder for digital transmissions, the steps of the method according to the invention.

It is specific subject-matter of the method according to the invention a memory medium readable by a computer, having a program stored on it, characterised in that the program is the computer program according to the invention.

It is further specific subject-matter of the present invention a system for the automated creation of personalized lists of channels for digital transmissions, comprising the use of a digital decoder associated to a player of digital transmissions, and a control of the decoder and/or the player, by one or more users, characterised in that:

- it comprises the use of a memory area in communication with said decoder;

- said decoder is provided with or connected to a computer means;

- said computer means are set up to execute the steps of the method according to the invention;

- said decoder is provided with means configured for the enjoyment of the transmissions according to one or more aggregation criteria on the basis of said reference vectors provided to the system or the vector associated to the current user in step F and/or one of the vectors calculated in step G according to the invention.

The invention will be no described by way of illustration but not by way of limitation, with particular reference to the drawings on the annexed figures, and in the specific illustrative case of terrestrial digital television transmissions, wherein:

- figure 1 shows an example of members constituting a family unit; - figure 2 shows a diagram of watching time-spans vector in the framework of the present invention;

- figure 3 shows a vector of watching time-spans aggregated by channel, according to the invention;

- figure 4 shows a vector of watching time-spans aggregated by genre, according to the invention;

- figure 5 shows a entity-relationship diagram that defines the relationships between member, user, family unit and class, according to the invention:

- figure 6 shows for example of family unit composed by five users (1-4 and generic) and by seven members (A-G), according to the invention:

- figure 7 shows an external of normalized vector of watching time- spans aggregated by channel, in the framework of the present invention;

- figure 8 shows an entity-relationship diagram of the user in the system according to the invention;

- figure 9 shows the dependency relationships between the various elements described in the entity-relationship represented in figure 8;

- figure 10 shows the functioning of the system according to the invention at the time of the first switching-on;

- figure 11 shows the functioning of the system according to the invention at generic time point T after the first switching-on;

- figure 12 shows the use of the vectors of the watching time-spans and watching time-spans aggregated by channel, according to the invention;

- figure 13 shows the logical scheme of the classification algorithm, according to the invention;

- figure 14 shows an example of relationships between classes in the framework of the present invention;

- figure 15 shows a mixed flow-blocks diagram of an embodiment of the system according to the invention;

- figure 16 shows a sample of affinity coefficients for a family unit with five users calculated by the system according to the invention.

Introduction

The present invention concerns a cognitive platform for the personalized enjoyment of multimedia contents coming from digital services such as for example DVB-T ("Digital Video Broadcasting- Terrestrial"), DVB-S ("Digital Video Broadcasting-Satellite"), and DAB ("Digital Video Broadcasting"). The basic characteristic of the invention is to allow an automated personalization of the receiver of the digital signal and its main functionalities without any human intervention. All the characteristics that provide a manual intervention are "additional" with respect to the main objective of allowing to the members of a family unit not to modify their own habits deriving from the use of the analogical TV, to the end of reducing as much as possible the uneasinesses in the passage from the analogical to the digital.

With the platform according to the invention, anyone will be able to continue to maintain the same habits acquired within analogical scenario. For example, in case of television multimedia contents, the user accustomed to the analogical world sits down in front of the television, switches it on by any button of the remote control and selects his own preferred channels by pushing in general an only button (or two ones at most) and interacts with a digital receiver by repeating the behavior acquired in the use of the analogical receiver. No complex manuals, complicated remote controls or new concepts to learn. In the specific case, the platform according to the invention would have as objective to allow each member of the family unit to maintain the same habits acquired in the analogical world even in the terrestrial digital scenario.

The reference scenario of the platform according to the invention provides a digital transmissions receiver, for example television (DVB-T and DVB-S) and/or audio (DAB) transmissions, utilized by a set of distinct physical persons. We will call such set of persons "family unit" and each single person "member". The digital receiver, containing the platform according to the invention, can be integrated in the apparatus (for example television or audio apparatus), or it can be external. Each member of the family unit can interact with the digital receiver remotely, for example by a remote control, or directly using the buttons on the interface of the same receiver. It is important to stress that the system according to the invention doesn't have the objective to identify the members of the family unit, rather it has the objective to identify the users of the family unit, i.e. to determine within the family unit one or more behaviors (of enjoyment of transmissions) that can be associated to a specific class. In the following, with the generic term "user", one defines one of said recognizable behaviors. To each user is associated an only one class, therefore the set of members representing such a class, from the point of view of the platform, behave as they were actually a unique constituent of the family unit. For example, an only class will be associated to two components of the family unit having the same likings and watching exactly the same transmissions and they will be considered by the system as a single user.

At the beginning, the user will be unique and representative of the whole family unit, however then, learning from the experience in a cognitive and dynamic manner, the system according to the invention individuates one or more behaviors that can be associated to one or more users classes.

One user class represents a category of persons within the nation

(or a region or alternative reference territorial entity) wherein the system of the invention operates. Example of class are: men graduated having age in the range of 42 to 52, undergraduate housewives, children below five years old, retired persons having lower level of education, teenagers, etc (the detail degree of the single classes can be more or less high).

By using data at disposal from different sources (Auditel, market analyses, etc) the invention associates to each users class preferably a pair of reference watching time-spans vectors: one with the time-spans aggregated by channels, another one with time-spans aggregated by genres (remark: the meaning and the motivation of the term "aggregated" will be introduced subsequently). One observes that, being these an illustration example of the television transmissions, it is here referred to "watching time-spans", however in other contexts one will have to use a corresponding terminology, in general they will be "enjoyment time-spans". Beyond making aggregation with respect to channels and genres, the aggregation of the enjoyment time-spans can be made by considering any other feature of the digital content, such as for example the author of the content, of the year of production of the content. We will call "aggregation criterion" any feature on the basis of which the enjoyment time-spans can be aggregated. Therefore, the reference vectors can be also more than two, corresponding to more than two aggregation criteria.

The reference vector of watching time-spans aggregated by channels is a vector (there is one for each class of users), constructed offline, i.e. not by the system according to the invention, having dimension equal to the amount of channels of digital receiver (usually this amount is 999, however we will indicate it by K) and that, in each element, gives the time-span that the components of a specific class of the user dedicate, within a predefined time window (for example one week, one month), to the watching of a specific channel.

This time-span can be referred to elementary time intervals on the basis of the needs of the market (for example, the format of the Auditel data in the television market), for example with respect to the timeslots, or single hour, etc.

Identically, the reference vector of watching time-spans aggregated by genres (genres whose number we indicate with M) in each element gives the time-span that the components of a specific class of the user dedicate, as observed in the predefined time interval (for example one week), to the watching of a specific genre. Even the single components of such vectors can be referred to elementary time intervals.

The definition of the classes and the relevant reference vectors is outside the objects of the platform according to the invention, however they represent a requirement for its correct functioning: indeed, such vectors will be provided as input to the classification algorithm of the system according to the invention that, without this information, is not able to classify the users. Hence, such vectors will be part of the configuration information of the platform according to the invention. It is important to remark that such vectors can vary along time, because national (or of the reference territorial entity) likings and habits can change along time, however the system according to the invention continues to function correctly without needing any modification, except the update of such input vectors.

It is important to note that the system according to the invention can function also in the case wherein the input information are not given directly as the above-mentioned normalized aggregated time-spans vectors, but in formats that can be connected back to the latter.

This result is achieved by the platform according to the invention by analyzing the behavior of the family unit, classifying the users (i.e. associating each user to a specific class) on the basis of the behaviors of the members of the family unit and by constructing personalized sortings (i.e. associated univocally to each user) of the channels: such personalized sortings are made accessible with the simplicity of the pressure of an only key (in the best case).

To the end of guaranteeing the privacy of the family unit, the platform according to the invention preferably does not carry out any type of identification: no personal information, identity information or in any case confidential information is necessarily requested to the members of the family unit. The system according to the invention is able only to classify the users and outline their profile. The profile of a user consists in maintaining a trace of the behaviors of subsets of members of the family unit, characterised by a similar modes of interaction with the digital multimedia contents. Even when an only member corresponds actually to a user, the system according to the invention is not able to detect in any way explicitly its identity, race, gender, age, religion, skin color or politics leanings. The system according to the invention, on the contrary, is explicitly able to classify and outline the profile of one or more users and to provide them a personalized enjoyment of the digital schedule, with the simplicity that is typical of the analogical scenario, on the only base of the channels most frequently watched.

To achieve such objects, the platform extends the functionalities of the today available receivers by adding the following functional features: basic functionalities:

o recording of the operations carried out by the remote control of the family unit;

o determination of the users configuration that best represents the family unit on the basis of the behavior classes manifested by the same family unit;

o determination of the current user on the basis of its use of the receiver;

o sorting of the channels on the basis of the user class;

o sorting of the channels on the basis of the user's preferences; advanced functionalities:

o dynamic classification of the genres preferred by the user on the basis of the behavior with respect to the current schedule; o composition of virtual channels (generic and/or thematic ones) created as a function of the profile and the user's preferences; o automated and personalized surfing of video channels.

The dynamic classification of genres is an advanced functionality because it can be carried out only in the case wherein information on the schedule are available. The electronic program guide (EPG), for example, provides some metadata useful to the dynamic classification of the genres preferred by the user. Preliminary definitions

By family unit (see figure 1), it is meant a set of distinct physical persons, who utilize the same digital receiver.

By member (see figure 1), it is meant a physical person that is part of the family unit that utilizes the decoder whereon the platform according to the invention is installed. A family unit is composed by at least a member.

By user it is meant instead a behavior that can be connected back (by an algorithm that is specified in the following) to a specific class of users. To each user one and only one class is associated. The number of users in a family unit is generally different from the number of its members. It can be smaller (if more than one member can be associated to the same user class), equal or larger (if a same member has behaviors that can be associated to different users classes or if there are aggregated behaviors between the users). A user is classified autonomously by the system according to the invention. An example of family unit composed by different users is shown in figure 6.

By default user it is meant the user that the system of the invention supposes present in the family unit at the first switching-on of the system. To this user, the sorting of the channels as provided by LCN is associated. As a matter of fact, it is the sorting of the channels that would be proposed to the default user in the absence of the functionalities of the system according to the invention.

The entity-relationship diagram defining the relationships between user, class, user behavior and class preference is given in figure 8.

According to the invention, a watching time intervals vectors (see figure 2) is a vector of dimensions equal to the number of the channels of a digital receiver. The value of each element of the vector is a list of watching time intervals relevant to the channel presenting the LCN equal to the progressive number of the same element. Each watching time interval is identified by a timestamp (i.e. by a watching starting time point) and a duration. Such information are obtained by capturing the commands that are given to the decoder, therefore in a vector no distinction is made between the various members of the family unit or users.

The watching time intervals vector is utilized for classifying the users of the family unit. Moreover, at each user of the system according to the invention is always associated the user behavior given by a user watching time intervals vector, as obtained starting from the watching time intervals vector. On the basis of aggregation criterion by channel and that by genres, one can obtain the aggregated watching times by channels and the aggregated watching times by genre, respectively. For each possible aggregation criterion, one may derive the aggregated watching times with respect to such a criterion. The system according to the invention extrapolates from the watching time intervals vector some synthetic information: the vector of the watching times aggregated by channels (see figure 3), the vector of the watching times aggregated by genre, the sorted list of channels and the sorted list of genres associated to that given user.

A vector of the watching times aggregated by channels (see figure 3) is a vector of dimensions equal to the number K of the channels of a decoder. The value of each element of the vector is given by the sum of the duration of the watching times of the same. Such vector will be defined in the same pre-defined time window for the reference vectors of the watching times and, like these ones, will be able to have a finer time granularity.

A vector of the watching times aggregated by genre (see figure 4) is a vector of dimensions equal to the number M of the genres supported by a decoder. The value of each element of this vector is given by the sum, on all the channels and all the watching time intervals recorded for each channel, of the values of suitable functions of the time points of watching start, durations of the watching time intervals and available information like EPG (for each channel and for each watching time interval). Such a vector would be defined in the same predefined time window for the reference vectors of the watching times and, like these ones, it will be able to have a finer time granularity.

The system according to the invention individuates, by an algorithm (users definition algorithm) and on the basis of the above-mentioned vectors, number and features of the users of the family unit.

In particular, the algorithm decomposes the vectors of the aggregated watching times into a suitable linear combination. The vector of the watching times aggregated by channel is decomposed into a linear combination of all the reference vectors of the watching times aggregated by channel of all the possible classes of users known to the system according to the invention. There could be existing a great many linear combinations that produce the same vector of the watching times aggregated by channel. The system seeks the most "scattered" solutions, that is among all admissible solutions the one is determined that has the largest number of components equal to 0 in the linear combination.

5 Analogously, the vector of the watching times aggregated by genre is decomposed into a linear combination of all the reference vectors of the watching times aggregated by genres of all the possible classes of users known to the system according to the invention. There could be existing a great many linear combinations that produce the same vector of the0 watching times aggregated by genre. In this case as well, the system seeks the most "scattered" solutions.

The number of users indicates the number of classes recognized within the family unit and is equal to the number of coefficients above a certain threshold of the linear combination. The value of this threshold5 represents a parameter of the system according to the invention and is instantiated off-line. Each user of the family unit is "characterised" by the value of the coefficient that multiplies the vector associated to a specific class (called coefficient of affinity to the users class). The coefficient of affinity has positive value: its value will be as much high as the identified o user approaches to the standard behavior of the class.

Once the number of the users, the classes to which they can be associated and the coefficient of affinity of each user to the class are known, the system has a series of objective data to recognize (by means of a suitable users recognition algorithm) the user when the latter begins to5 watch the television programs. Thanks to the recognition users algorithm, the system is able to make an assumption on the class to which the current user belongs.

When the current user is recognized, the system according to the invention assigns to it a user and, on the basis of the channels watched by o this user, updates the vector of the watching time intervals of the user.

After a period of recording of the user behavior (for example one week), the vector of the watching time intervals of the user can be used to calculate the vector of the watching times aggregated by channel and/or genre of the user, which will be compared to the homonymous vectors5 relevant to the class associated to the user (multiplied by a coefficient of affinity as defined in the following). A significant value of the distance (according to suitable distance functions, that can be chosen to improve the recognition of the users) between the two vectors will indicate the necessity of a new execution of the user definition algorithm. On the contrary, reasonably small values of the distance between the two vectors will indicate a stable identification of the user. This mechanism of feedback and correction of the error contributes to stabilize the process of learning.

When the system is not able to identify the current user (for example to the first switching-on, or after a reset) all the recorded data are associated to the default user. Such a user is the one characterised by preferences typical of the average user (or equivalent), i.e., in the national case, corresponding to the national classification of the audiovisual digital channels (sorted on the basis of the LCN value, as explained above).

By class preferences it is meant a pair (or sets if the aggregation criteria are multiple) composed by a list of N sorted channels, corresponding to the channels preferred by the members of such class, and a list of M sorted genres, corresponding to the genres preferred by the members of the same class. Obviously, the number N of preferred channels can be at most equal to K (total number of channels that are manageable by the system) and corresponds to the number of significant positions in the list of the K available channels sorted by decreasing preference level. Equivalently, the number M of preferred genres can be at most equal to the number G of defined genres and corresponds to the number of significant positions in the list of the G defined genres, sorted by decreasing preference level.

By classification of the family unit, it is meant the automatic determination of a finite number of users by the platform according to the invention starting from the behavior of the members of the same family unit. At the beginning, the system according to the invention does not have sufficient information for classifying the family unit and therefore has only a default user, classified as belonging the default class. As the system according to the invention learns, it can determine one or more users, on the basis of the use of the receiver by the members of the family unit.

As one sees in figure 8, to a user is associated one and only one class (and, vice versa, to each class corresponds, in the family unit, one and only one user) and an only user behavior. To each reference class, one and only one class preference is associated, given by a pair of sorted lists, one of genres and one of channels. To the user behavior, instead, is associated a vector of watching time intervals, from which one can obtain all the other vectors.

It is to be considered that, starting from the user behavior (i.e. from the vector of the watching time intervals) one obtains its classification (i.e. the membership to any only given class). It is therefore possible, starting from the only user behavior, to associate to the user himself the lists of sorted channels and genres. To each classified user, indeed, the system according to the invention can suggest channels (all those given in the sorted list of the channels associated to the class preference) or genres (all those given in the sorted list of the genres associated to the class preference). Besides, the system according to the invention can also elaborate, on the basis of the user behavior, i.e. the vector of the watching times of the recognized user, a list of genres preferred by the user by sorting the channels on the basis of the vector of the watching times aggregated by channel.

In figure 9, the dependency relationships between the user, its membership class and different types of vectors utilized by the system are shown. The system according to the invention records, for each user, a vector of the watching time intervals. From such a vector, one can derive the vector of the watching times aggregated by channel. From the vector of the watching times aggregated by channel, the system according to the invention can individuates a sorted list of channels preferred by the user. From the vector of the watching time intervals, one can derive also the vector of the watching times aggregated by genre if information are at disposal concerning the genres and relevant to such watching time intervals (for example by means of the Electronic Program Guide). From the vector of the watching times aggregated by genre, the system can individuate a sorted list of genres preferred by the user. The classification of the user is a dynamic and continuous process. The user class allows the system according to the invention to associate to the user a sorted list of channels and a sorted list of references genres, i.e. the sorted lists of the corresponding class preference (see figure 8).

Functioning modes

The system according to the invention starts acting as soon as the decoder or the device whereon it functions is switched on. Making reference to figure 10, at the first switching-on, the system according to the invention assigns to the current user the default class to which the sorted list of channels corresponds, that is given by the national standard numbering of channels LCN (position 1 - channel with LCN 1 , position 2 - channel with LCN 2, position N - channel with LCN N). Moreover, the system according to the invention initializes the vector of the watching times. The current user to whom the default class is assigned is called generic user.

Starting from the vector of the watching time intervals (which is initially empty), the sorted list of channels of the user and the sorted list of the genres of the user (they also initially empty) are realized, which are indicative of the generic user's behavior. It is to be observed that, besides the channels and genres, different criteria of aggregation of the digital transmissions (for example, the author of the transmission, the production year, etc) are possible. For each criterion, it is possible a classification of the population into several classes of users. For each criterion of aggregation available for the definition of the users classes, the system acquires the corresponding class sorted list and initializes the corresponding user sorted list: the first defines the sorting according to the class preferences; the second one defines the sorting according to the behavior of the single user. As the decoder is utilized, the watching time intervals of the single channels are memorized in the vector of the watching time intervals associated to the generic user, characterizing therefore its behavior along time (see figure 11). In this step, by pushing a dedicated button (for example the "0") the system user (who, let's repeat it, for the time being is still identified as "generic user") accesses the preferred channels, i.e. the channels as in the user channels sorted list (see figure 11). In other words, following the pushing of the dedicated button, the system according to the invention allows the visualization of the first channel in the user channels sorted list. With the buttons "+" and "-" of the remote control or the usual numerical keys (for example, 1 for the first of the list, 2 for the second and so on), the user can move within the user channels sorted list (i.e. between its preferred channels).

Following further pushing of the dedicated button, possibly different from that indicated in the previous case, the system user accesses the preferred genres, i.e. the user genres sorted list, with a functioning similar to the previous one: to the pressure of the key "1" corresponds the watching of a digital transmission whose genre corresponds to the first genre in the list, and so on. In the case that for a given genre more than one digital transmission are present in the same genre, it is possible to adopt one or more selection criteria provided by the system according to the invention. For example, the automated and personalized channel surfing or the digital channel according to the invention.

By pushing a dedicated button, possibly, in turn, different from those indicated in the previous cases, the user accesses the current user class. Being in this case (first switching-on) the current user equal to the generic user, as above said, the class channels sorted list corresponds to the national standard structure of the LCN channels. Like for the other lists, the user can move within the list with the keys "+" and "-" or with the normal numerical keys.

If this is available, besides channels and genres, any other type of criteria of aggregation of the digital transmissions (for example the author of the transmission, the production year etc), to each new pushing of the dedicated button, the user accesses and new list sorted according to another aggregation criterion. Indeed, for each aggregation criterion available for the definition of the users classes, the system makes the corresponding class sorted list available to the user. For the sake of simplicity, in the figures 10 and 1 1 , the user sorted list and the class sorted list relevant to the only criterion of aggregation by channels are given.

Once terminated the set of available aggregation criteria, the further pushing of the dedicated button allows the user to go back to the initial watching mode, corresponding to the sorting by channels according to the national standard sorting of the LCN channels.

The criterion for choosing the keys on the remote control to be associated to the above-mentioned commands, which is not part of the claimed features, should take into account a twofold aspect: on one hand, it should allow the user to utilize dedicated keys that are also easily identifiable on the remote control; on the other hand, it should reuse those keys that the user is already accustomed to use to change the TV functioning mode (for example, depending on the TV model, the key "AV", the key "Source", ...).

For a better clarity, it is here specified that according to the embodiment here described, the digital receiver memorizes four different lists: the user channels sorted list of length N; the class channels sorted list of length M; the user genre sorted list of length L; the class genres sorted list of length P. The system initializes and manages moreover the vector of the watching time intervals of the user which memorizes the watching time intervals for each one of the K digital channels provided by the national standard sorting of the channels LCN. Without any loss of generality, in figure 10 and 1 1 there are shown, respectively, at the time of the first switching-on plus a certain time, both the vector of the watching time intervals of the generic user and the sorted list of class and the sorted list of user with respect to the only criterion of aggregation given by the channels.

If any other type of criterion of aggregation of the digital transmissions (for example, the author of the transmission, the production year etc) is available, besides channels and genres, for each aggregation criterion X, the ordered list of channels of the user having length Νχ and the sorted list of channels of class having length Μχ are defined. In the case of the generic user, the sorted list of channels of the default class corresponds to the subset of the first N elements of the national standard sorting of the LCN channels including the map of all the possible K DVB-T channels accessible by the receiver.

From the first switching-on of the system on (and therefore at a given time instant T), the system executes the algorithm of automatic classification of the users. The objective for such an algorithm is to decompose the user behaviors known to the system at the time point T into Y (with Y a positive integer) distinct user behaviors. Y is the number of distinct users, each corresponding to a different reference users class, individuated by the algorithm of classification within the family unit.

Starting from the vector of the watching time intervals, the vector of the watching times aggregated by genre and/or the vector of the watching times aggregated by channel are calculated. In general, one will have a vector of the watching times aggregated by each criterion of aggregation of digital transmissions. The algorithm of classification works always on an only vector at the time and in the following reference will be made to it with the name of vector v. By suitable algorithms, the vector v is decomposed into a linear combination of reference vectors of the watching times aggregated by channel and/or by genre v, associated to the classes / = 1, ...,V of standard users. Let's define V = {vy , v ₂ , v _y } the set of all the vectors associated to the classes of standard users (as defined on the basis of the Auditel data or added during the functioning of the system according to the invention), the algorithm determines the coefficients a, (affinity coefficients) of the following linear combination :

The subset Y of V corresponding to the coefficients a, larger than a predefined threshold (significant affinity coefficient) constitutes the new set of the users individuated by the system according to the invention within the family unit.

It is to be stressed that the linear combination

V =∑ieV Cli Vi

Exists only if the set V of reference vectors contains a subset of Y vectors that constitute a base for the vectors space having a number of components equal to the number of channels.

In order to guarantee the fulfillment of this condition, the system according to the invention completes always the set V by introducing suitable pre-defined "fictitious users" corresponding to vectors of watching times having an only component (for example, one hour for a specific channel) put e.g. equal to 1 (and the other components equal to 0). In other words, by interpreting such a mathematical completion, the "fictitious user" is, according to an aspect of the invention, a user which follows an only channel within a specific time interval (one hour in this example). The reference vectors of the watching times aggregated, defined for the fictitious users allow to complete the base formed by the vectors with vectors u _z corresponding to Z fictitious users. Hence, the system according to the invention is always able to determine a (admissible) solution of the system: V =∑ieV a\ V/+∑ eZ ¾ «/

The term bj Uj represents the deviation with respect to the decomposition of the family unit as a linear combination of the standard users classes weighted by suitable affinity coefficients. If the number of non-zero components of the vector v° (number of overall watching hours of the "fictitious users", in our example) comes out to be smaller than a predefined threshold (for example less than 5% of the overall hours of the time interval chosen beforehand), then the above illustrated set Y (corresponding to the set of standard users with affinity coefficients larger than the significant coefficient of affinity) can be considered an acceptable approximation of the composition of the family unit. If instead the number of components different from zero of the vector v° exceeds a predefined threshold, then the deviation can be interpreted as the presence of a member of the family unit that is not comprised in any of the classes given as input to the system according to the invention.

In this case, the recognition algorithm will create a new user of the family unit with vector of watching times and a new standard class obtained by inserting the vector Vo in the set V. This new user will be able to be recognized by the system and managed exactly the like standard users, given as input to the system according to the invention. The list of his preferred channels will not be decided a priori, as it happens for the standard users, but it will be defined by the system according to the invention as the new user is identified and his sorted list of the channels of the user is updated.

Coming back to the above example (that we recalled to be started the from the first switching-on), as shown in figure 1 1 , as soon as a suitable quantity of information will be collected and analyzed, which concern the use of the decoder whereon the system according to the invention is installed and, therefore, to a certain time point T, the automatic users classification algorithm will be activated, which will be applied to the only available vector of the aggregated watching times: the one of the generic user. From the vector of the watching time intervals, one will be able to calculate the vector of the aggregated watching times, as obtained by simply summing up the various time intervals for each channel (see figure 12). To the so obtained vector of the aggregated watching times, the classification algorithm is applied, which takes as input exactly such a vector and suitably divides it into Y new vectors as above illustrated.

Assuming for example that the family uniti s composed by two persons and that the digital transmission system is constituted by ten channels, and that five classes are available with the relevant reference vector set V={v _{l t} v ₂ , v ₃ , v ₄ , v ₅ ) (aggregation criterion by channels in this example) and the set U={U], U2, U3, u ₄ , u ₅ , u ₆ , u ₇ , u$, Ug, u ₁₀ } of aggregated watching times vectors relevant to ten fictitious users, each of them watching one of the ten channels for exactly a hour. As a consequence, U will have the first component equal to 1 and the other components equal to zero, U2 only the second componente qual to 1 and so on. The classification algorithm takes as input a vector v of ten components, each of which represents the watching time of the family unit aggregated by channel, and determines a solution of the (admissible) system of ten equations and fifteen unknown quantities (ai, α∑, a^ a ₄ , as e bi, b2, bio):

At this point there can be infinite solutions of such an equations system, and the method according to the invention in such a case seeks an optimal solution. For example, the optimal solution can be obtained by a mathematical programming method, and can be that with the largest number of components equal to zero, in such a way to have a sparse solution with respect to the reference vectors.

Once obtained the optimal (or sub-optimal) values of the coefficients a/, ^, 03, a ₄ , as, the coefficients whose optimal value is below a predetermined threshold indicate which vectors of the set V are to be inserted in the set Y.

The result of the algorithm in terms of classification of the family unit is to formulate, each time that a suitable quantity of information has been collected and analyzed, which concern the use of the decoder, a hypothesis of composition (linear combination of reference classes represented by users belonging to the family unit) more accurate it than that known till time point T. Each time that this hypothesis of composition is updated, the system assigns to a new user a suitable reference class. As shown in figure 13, a possible example of classification comprises the decomposition of the vector of the watching times aggregated by channel of the generic user and its classification into three users. The first user is associated to the student class, the second user is associated to the retired persons class, the second one is associated to the minors class (more generic). It is noted that in this example, starting from time point T, three users exist of which two ones are totally new with respect to the generic user (the only one recognized by the system till that moment). The two new users are well classified, in the sense that they are recognized as users of known reference classes (students and retired persons), whilst the generic user is passed from the default class to a class that is less generic but still uncertain since it approaches a class containing for example all the minors.

Each classified user, by pushing a key dedicated to a suitable combination of keys, accesses the sorted list of user's channels, or the sorted list of user's genres, or the sorted list of class channels corresponding to his own reference class, or the sorted list of class genres corresponding to his own reference class. It is here recalled that, whenever one or more criteria of aggregation of the digital transmissions are available besides channels and genres, by pushing a dedicated key, the user accesses the corresponding class sorted list and user sorted list. The classified user, by pushing a dedicated key or a suitable combination of keys accesses therefore again the standard national sorting LCN of the channels containing the map of all the K DVB-T channels accessible to the receiver.

After the time instant T, the digital receiver can be switched on and switched off several times. At each new switching-on, the system according to the invention always assumes that the user who is using currently the receiver is a determined user, according to a determined policy of realization of the system. Such a predefined assumption can be wrong, because made a priori, and the system operates therefore in a user monitoring mode. In such a mode, the system applies the recognition algorithm and verifies to which user corresponds the minimum value of the sum of the positions according to the classification of the programs watched in the first scan of the programs carried out by the user of the system. The system according to the invention, by observing, by means of the recognition algorithm, the behavior of the user who is currently using the receiver, can make an assumption, and associate, instead of the "predetermined" user (for example, the generic user), one among the users classified until that moment, in particular the one to whom the current behavior of the receiver user best fits.

After the instant T, a new time instant, which is here called T', can make another process of classification begin; by continuing with the timeline of the foregoing example, because there are now more than one classified user, the following procedure will be applied:

1. All the vectors of the watching times aggregated by channel (or another available criterion of aggregation of the digital transmissions), are summed up to obtain an only overall vector of the whole family unit;

To this vector, the classification algorithm is applied;

The obtained result determines the new configuration of the classified users;

If the configuration is equal to the previous one, nothing happens. In the opposite case, the new classification is activated by the identification algorithm. The classes

The classes of the system according to the invention are needed to collect, according to the criteria of aggregation of the digital transmissions (such as for example those typical of the marketing), suitable subsets of the national population. The criteria of aggregation of digital transmissions and determination of the classes are rather complex. To understand the functioning of the system according to the invention, an example will be proposed to introduce the concept of relationships between classes. The national population can be indeed gathered with very general classification criteria or with increasingly detailed criteria. For example, the set of all the persons of the nation can be classified by the dichotomic separation criteria realized by means of questions (as it happens in a decision tree). As one provides an answer to a question, one identifies subsets of the population, and therefore classes of users, more and more detailed.

Let's take as an example figure 14. In this figure, K classification criteria are represented: sex, age, graduation, etc. Such K binary classification criteria identifies 2 ^{(K+ )} -1 classes on the whole. The first class is the default class and gathers all the possible the citizens of a nation. Class 2 represents all the male citizens, whilst class 3 represents all the female citizens. Class 4 represents the graduated male citizens, and so on. Obviously, as one goes down in the classes tree, one tends to typify and better characterize the peculiarities of a class, till one arrives to determined leaf classes, which are highly specialized classes like, for example, housewives, students, infants, retired persons, etc.

The digital receiver according to the invention applies a classification criterion starting from a set of classes with relationships as exemplified in figure 14. The set of classes and their relationships influence obviously the classification algorithm and, depending on the number, type of classes and their reciprocal relationships, which are configured as data for the algorithm, one will obtain extremely heterogeneous functionings. The system according to the invention will not tie to a specific set of classes, it provides instead the possibility to define them at will and a priori, but also to update them in the lifetime of the receiver whenever the behaviors of a member of the family unit is not comprised in any reference class. This obviously renders possible the personalization of the digital receiver to the end of carrying out different marketing strategies, as a function of the sets of classes that one each time wishes to utilize, to perform, for example, targeted, temporary or simply experimental marketing campaigns.

In figure 15 one distinguishes the standard functionalities (colored in white) and the advanced ones, due to the system according to the invention (colored in grey), which are subject matter of the present patent application (not all the connections between the components are indicated, but only the most significant ones). The system according to the invention is constituted by interoperating functional blocks that are called "engines". Each engine performs specific functions of the platform according to the invention and their interoperation allows to reach the preset objectives.

We analyze in detail the functionalities of each engine (software modules).

- CONFIGURATION ENGINE. This subset deals with the management of all the configuration information of the system according to the invention. By suitably interfacing with the subsystem of remote connection, the configuration engine can receive for example updates on the classes, on the corresponding normalized vectors of watching times aggregated by channels and by genres and on the national classes preferences, and can record them on its database. All the configuration information, comprising all the parameters of the system according to the invention, are made available by the configuration engine to the remainder of the architecture according to the invention.

- INTERFACE ENGINE. Such subsystem deals with intercepting the commands sent to the decoder through the remote control or the possible keys on the same decoder, and to send them to the monitoring engine to record them, and to the rule engine to know which command is to be sent actually to the decoder. The interface engine deals with actuating the decisions of the rule engine as well, which can provide an interaction with the OSD (On-Screen Display) of the device.

MONITORING ENGINE. It deals with recording in its own database all the indirection operations carried out with the decoder to the end of keeping trace of channels actually visualized and therefore to determine the vector of the watching times of the family unit.

CLASSIFICATION ENGINE. It deals with classifying the family unit on the basis of the data provided by the monitoring engine. The classification engine works constantly in background to determine the configuration of users, among those present in the configuration engine, such to best represent the behavior of the family unit. The users configurations as determined by the classification engine are recorded in the relevant database. The classification engine deals with determining, on the basis of the current interaction with the decoder, which user is actually using it on the basis of the previously described monitoring, as well. Indeed, the module implements also the algorithm of user recognition of the system according to the invention. On the basis of the family unit classification, the module produces therefore an assumption on the membership class of the current user.

PROFILE ENGINE. It deals with keeping trace of class and user preferences for each single user. Once the configuration of the users and the current user using the decoder are known, the profile engine updates constantly the users profile on the basis of the information as determined by the monitoring engine. All the information are therefore recorded on the internal database.

RULE ENGINE. As a function of the command intercepted by the interface engine, the rule engine deals with determining the real action to be initiated and therefore the possible command(s) to be sent to the decoder. The actual action corresponding to a command sent by the user to the decoder is determined as a function of the users profile and the status of the decoder. Hence, the rule engine interfaces both with the interface engine and the profile engine. EPG ENGINE. This module deals with extracting EPG ("Electronic Program Guide") information, both from the DVB stream and possibly having resort to information retrieved on line, useful to the determination of the genres associated to the different channels. The information of the EPG engine are therefore utilized by the profile engine to establish the genre preferences of the user.

Virtual channels

The concept of virtual channel is very simple and is based on elementary observation: each user, during the day, passes his time before the television means, searching transmissions and programming or, depending on the hour of the day, going directly and without fail to a given channel, or searching a transmission of a genre congenial to him among a subset of channels that he knows, i.e. those that the believes transmitting in that given moment something that he can appreciate. Finally, if any of the two cases does not satisfy his search, either he goes through all the channels till he finds the programming best congenial to him, or he watches one of his "preferred" channels waiting for a future transmission.

The user behavior during the search, therefore, is quite repetitive and, in any case, follows easily identifiable patterns. In such a connection, it is highly likely that one is able to determine automatically, at each moment of the day, which transmissions (and therefore which channels) could be appreciated by the user in that given instant.

The system according to the invention allows to recognize the single user of a family unit by assigning to each of them both an ordered list of class channels and an ordered the list of user channels. If other criteria of aggregation are at disposal, to each user of the family unit also the ordered class and user list is are associated, which is relevant to each aggregation criterion (for example, genres, authors, year of production, etc). As a consequence, once the person utilizing the system is classified and recognized as a user, one can realize two solutions: the first one is simply to realize "pop-ups" which indicate to the person utilizing the system the beginning of a transmission which could be interesting for him on the basis of one of the above ordered lists; the second one provided the creation of virtual channels.

The first one, in particular, presents some problems: first of all it could "disorientate" persons utilizing the system which are not used to provide a feedback to the decoder. Moreover, the number of pop-ups could be exceedingly high or in any case a pop-up in the middle of a hot point of a transmission could be particularly unpleasant. The second solution solves these problems. The basic idea is to map on "fictitious" channels (for example channel 12345, to which nowadays no LCN channel corresponds) the programming of other real channels, sorting them chronologically in such a way that the stream be most possibly continuous and/or corresponding to the user preferences, and differentiating the various virtual channels into "thematic" and "general" channels, depending on the fact that transmissions follow one another which share the same genre or not.

To make example, on channel 10001 , which could be general, from

00:00 to 01 :20 the transmission of the LCN channel equal to 1 could be mapped; subsequently (01 :20 - 03:05) the transmission of the LCN channel 32; then (03:05 - 03:30) that of LCN 3 and so on. On the channel 20001 , which would be a thematic channel, from 0:00 to 00:30 the transmission of LCN 12 channel could be mapped, from 00:30 o 02:40 that of LCN 2 channel and so on. The difference, in the second case, is that all the transmissions of the thematic virtual channel share the same genre/theme.

The advantage of this second solution is that first of all there would be no pop-ups and interaction: at each given moment of the day, among general channels there would be, in order strictly decreasing with channel, the transmissions likely less interesting for the user (from 10:00 on 10001 , there would be the transmission that the system has identified as the most interesting one, from the point of view of the user, in that moment; at the same hour, on 10002 the second most interesting one: on 10003 the third most interesting one and so on), whilst on the thematic channels there would be, instant by instant, (if possible) the transmission most interesting for the user in that moment and of that given genre (if the aggregation criterion by genre is available). In particular, in this second case, the user could utilize the keys "+" and "-" of the remote control, however not for changing genre (i.e. passing from channel 20001 to channel 20002) but, as though remaining on such channel, to pass to the second transmission of that genre that is present in that moment. Other advantages of this solution are that the user will be able to practically remain (ideally) on a single channel.

Automated and personalized channel surfing The concept of channel surfing is very simple: when a member of the family unit accesses the digital receiver or has already in mind what is to be watched, or he performs a more or less random search aimed at finding an interesting program. The last behavior is called channel surfing, i.e. the person utilizing the receiver passes rapidly from a channel to another one, unless he stops when he finds something interesting. A minimum watching time is necessary to realize whether the programming content currently on the channel is interesting or not (for example L seconds, with L suitably configurable).

The personalized channel surfing has the objective to perform automatically the channel surfing on behalf of the current user. When the user decides to perform the automated channel surfing, the system according to the invention selects on the basis of the user preferences, its class, the day of the week and the timeslot and (in an advanced version of the service) also the basis of the user preferred genres, those channels which could best interest the user and visualizes them one after the other for L seconds each. When the user finds an interesting channel, it stops the automated and personalized channel surfing offered by the system according to the invention and enjoys its contents.

Once the automated and personalized channel surfing service is activated (if for example the user is not interested to some proposed channel, believes that too long a time is to be waited for the passage between a channel and the subsequent one, or he was late to block the channel that was drawing his attention), the user can, by pushing keys "+" and manually force the scroll or go back in the suggested channel list.

In the foregoing, embodiments have been described and variations to the present invention have been suggested, it is however to be understood that those skilled in the art can modify and change such embodiments, without departing from the relevant scope of protection, as defined by the attached claims.