FIELD OF THE INVENTION
The present invention relates to generation of multimedia recommendations to a user in
general and to a method and system for generating multiple users' recommended visual
precursors of multimedia contents in consumer electronics devices such as Digital
television, handheld devices connected with internet in particular.
The invention relates to a method for generating a suitable visual precursor or a sample
to be shown to a viewer from a clip that is recommended on his/her computing device
which is connected to internet. When a user watches multimedia content on a digital
television or for that matter any CE device that is connected to internet, he / she can be
recommended other multimedia contents based on his / her preferences. The multimedia
content selected for recommendation could be selected by filtering techniques such as
collaborative filtering or social networking. This invention focuses on determining which
parts or scenes of the recommended multimedia content should be shown to the viewer
that can generate his / her interest of watching the multimedia content on the CE device.

BACKGROUND AND PRIOR ART OF THE INVENTION
The utilization of sociological networking is progressing in the mass-media world. Thus,
an analysis of existing recommender system in consumer electronic devices such as
television was done. The current work for micro-personalized recommendation for
multimedia content which clusters the population into groups according to users'
preferences, socio-cultural trends, and homogenous behaviors was examined.
Consequently, it was established that recommender systems research that has focused
on the interaction between information retrieval and user modeling can be further refined
in order to provide a more personalized, custom-made and proactive retrieval experience
and help users choose between retrievals on his/her consumer electronic device which is
connected to internet either wired or wirelessly.
The present invention provides a method and system, which extracts scene segments
from the viewed content by other users. The additional user preference data helps in
refining users preference matching. The present invention provides a framework capable
of presenting recommended content visibility by showing the scene segments collected
from other users who have rated/watched the same multimedia content.

US patent publication 2006/0174260 Al provides a method for generating
recommendations, the method including: prompting a user for feedback on at least one
preference for generating a recommendation, the at least one preference having two or
more categories associated therewith; displaying at least one visual cue corresponding to
each of the two or more categories; selecting one of the two or more categories based at
least in part on the corresponding at least one visual cue; and generating a
recommendation based at least in part on the selecting. Preferably, the generating
generates a recommendation for video content, such as a television program and the at
least one preference is the genre of the television program, such as action, drama,
comedy-action, suspense-action, comedy, documentary, or romance.
US 6,453,471 discloses a preview system activated from an on-screen programming
guide displays a video preview of a selected particular program on the display screen.
The video preview is displayed on less than the full screen so that the video can be
highly-compressed to save bandwidth.

JP 2006054630 A teaches a program recommendation method which includes the steps
of video-recording or audio-recording a transmitted program; and recommending viewing
of a plurality of scenes configuring the video-recorded or audio-recorded program by
means of weighting according to the program rating of the scenes. It discloses a
program recommended practice including a process of recommending viewing and
listening of two or more scenes, which constitute said program recorded or recorded by
weighting according to a size of viewership of the scene concerned.
The present invention seeks to overcome the above mentioned limitations of the prior
art.
SUMMARY OF THE INVENTION
Current recommender systems in the consumer electronic devices such as digital
television, and electronic devices connected to internet implement a rather limited model
for multimedia content visibility based on filtering techniques. Most of these systems fall
short to provide visual precursor to the user and concentrate only on making more

accurate predictions; however, a few of them that focus their attention to the aspect
of multimedia item visibility do so in a limited scope. In this invention, we address this
problem and propose to augment the existing recommender systems with a dynamic
user-based scheme to provide users with superior, high-quality recommendation
formulation and customized visibility of the recommended item in the consumer electronic
devices. The domain of content visibility is dynamically crafted using the existing
recommender system algorithm. This is performed by summing the ratings of various
scenes assigned by other viewers, weighted by the similarity of the taste of the individual
other users with the target user. This not only ensures that the multimedia content item
suggested are relevant to the user, but also that he / she is shown the parts that are
most likely to attract his / her attention.
The invention provides a method for generating multiple users recommended visual
precursors comprising of the steps of: demarcation and selection of the scene segments
by the user using his CE device capability; inputting rating information for a particular
scene segment in the recommender system; asking for N best recommendations from the
recommender system; injecting the selected scene segments one by one into scene
space by the scene patch, and generating the complete visual pre-cursor shot,
characterized in that the complete visual pre-cursor generated by the method is matched
to the taste of the user and is personalized.

BREIF DESCRIPTION OF THE ACCOMPANYING DRAWINGS
The invention can now be described in detail with the help of the figures of the
accompanying drawings in which
Figure 1 shows the generation of visual pre-cursor shot from all scene patches.
Figure 2 shows the determination of video rating by weight-time graph method.
Figure 3 shows functioning of the recommender system for video request by a user.
Figure 4 shows functioning of the recommender when N
best recommendations sought.
Figure 5 shows an exemplary scene generation.
Figure 6 shows brief system architecture
An exemplary and non-limiting embodiment of the invention will now be described with
the help of the accompanying drawings.

PESCRIPTION OF THE INVENTION
Rapidly evolving consumer electronics market, tied with the exponential growth of the
video-on-demand services and information available on the internet, have already
brought businesses and users to the point where hundreds of millions of users seek fast,
pervasive access to phenomenal amounts of multimedia content. The challenge of
complex environments is therefore noticeable: recommender systems are expected to
perform optimally well in different situations for a variety of users and for a variety of
communications. To cope with such situations, the promise of smarter recommendations
is becoming highly attractive.
As shown in Figure 6 the Internet based recommender engine 101 broadly comprises
interaction with a plurality of users 102 whose scene preferences are stored in user
preference database 100. The consumer electronic device 104 has a display module 105
in communication with the recommender system wherein the recommender system
interacts with the recommendation engine via internet 104 to show the recommendation
multimedia content items to the user. The electronic device may be controlled by remote
control 106.

In current scenario, recommender systems development and improvement focus
attention to make better predictions using past user's behaviour pattern and are
employed to personalize services on the internet as well as in electronics devices, such as
television. These systems on consumer electronic devices, which are connected to
internet, learn from user, based on rating feedback from the user on, for example, songs,
movies or television programs.
However, these recommender systems address and focus more on filtering algorithm
issues, and are therefore unable to leverage the benefits from content visibility. Content
visibility is as important as any other technical issue, and may indeed have a large
positive impact on the usability and performance of these systems. For example, viewing
users' pattern and provision of solitary content recommendation does not address the
problem of false positive recommendations.
The recommender systems do not exploit the fact that users tend to like particular
segments of the viewed content more than the rest. The extraction of these segments
and their information can further enrich user experience, improve quality viewing and
advance recommender systems.

To address this belief, we put forward in the present invention a novel dynamic user
based scheme in the consumer electronics devices. The scheme facilitates the formation
and maintenance of existing social networking architecture, and capitalizes its social trend
for improved content recommendation, and sharing. It further helps in providing
personalization and building a sophisticated recommender systems that track and
recognize users' segment preferences in the viewed content in the CE devices.
The video-on-demand based services on the television stream telecast the multimedia
content. The systems can be associated with powerful recommender systems, which can
suggest multimedia (video) or predict taste of certain multimedia for a particular
television user and opinions of other like-minded television users. The weakness of the
CF algorithms for effective visual pre-cursor led us to explore improvement in multimedia
recommendations of these systems. After screening and reviewing several techniques,
we decided to develop a scene matrix generation (SMG) algorithm to improve the
dimensionality of the current recommender system.
More formally, we assume the presence of a set of television users, TVU = {tvu1, tvu2,...,
tvum}, and a set of multimedia (video) items, Ml = {mi1, mi2,..., min}.

The scene matrix for a television user tvueTVU can be viewed as an n-dimensional vector
of ordered pairs,

In recommender systems in the electronic devices involving scene ratings by the
television user for multimedia items, the function ptvu is called a scene rating function,
mapping the multimedia items to a discrete set of scene ratings. The mapping ptvu for a
given jth television user tvuj is associated on the whole section of multimedia items, the
recommender system records the interested scene of a given television user for
multimedia content. Scene matrix for a particular multimedia item is collected over time
and stored by the recommender system.
The algorithm developed for the consumer electronics devices for the generating SMG
has an operation in which it atomically outputs contents visibility scene to a television
user.
SMG, which uses television user's scene taste input as its underlying dimensionality
improvement, maps nicely into the collaborative filtering recommender algorithm.

WE CLAIM
1. A method for generating multiple users recommended visual precursors of a
multimedia content comprising of the steps of:
- demarcation and selection of the scene segments by the user using his/her
consumer electronic device;
- inputting rating information for a particular scene segment in the
recommender system using his/her consumer electronic device;
- asking for N best recommendations from the recommender system;
- injecting the selected scene segments one by one into scene space by the
scene patch, and generating the complete visual pre-cursor shot,
characterized in that the complete visual pre-cursor generated by the method is
matched to the taste of the user and is personalized.
2. The method as claimed in claim 1, wherein selection of the scene segments may
optionally incorporate ratings associated with the particular scene of the
multimedia content item.

3. The method as claimed in claims 1 and 2, wherein the method can be applied to
video-on-demand services provided on electronic devices.
4. The method as claimed in claims 1 and 2, wherein the multimedia content is a
television program.
5. A method for generating recommendations, the method comprising:

- prompting a user for feedback on scene preferences for generating a
recommendation, the at least one scene preference therewith;
- displaying visual precursor corresponding to the collection of scene
preferences of other users by the system described; and
- generating a visual precursor of the selected scenes for the
recommendation / recommended content.
6. The method as claimed in claim 5, wherein the recommender system on the
electronic device connected to the network generates a recommendation and its
associated visual precursor for multimedia content.

7. The method as claimed in claim 5, wherein the displaying comprises:
- displaying means for scene selection to correspond to the favorite/liked
scenes in the multimedia content; and
- displaying visual precursor corresponding to scene selection of the selection
means to the user.
8. The method as claimed in claim 5, wherein the displaying of the selection-means
comprises the following;
providing a user interface having a button to display proximate scenes of the
multimedia content item wherein the selection of the scenes is achieved through
the user interface on the electronic device, which is connected to the network.
9. An apparatus for generating visual pre-cursor recommendations, the apparatus
comprising:
- means for creating a user interface on the electronic device, which is
connected to the network, for prompting a user for scene likeness selection;
- means for displaying visual pre-cursor corresponding to scenes selected by
multiple users to the user who has been recommended a multimedia
content; and
- a recommender for generating a visual pre-cursor recommendation based
at least in part on the scene selection.

10. The apparatus as claimed in claim 9, wherein the recommender generates a
recommendation for video content.
11. The apparatus as claimed in claim 10, wherein the multimedia content is a
television program.
12. The method as claimed in claims 1 to 3, wherein determination of multimedia
rating is done by weight-time graph method.

A method for generating multiple users recommended visual precursors comprising of the
steps of: demarcation and selection of the scene segments by the user; inputting rating
information for a particular scene segment in the recommender system; asking for N best
recommendations from the recommender system in the consumer electronic
device/devices; injecting the selected scene segments one by one into scene space by
the scene patch, and generating the complete visual pre-cursor shot, characterized in
that the complete visual pre-cursor generated by the method is matched to the taste of
the user and is personalized.