System for hardware, channel, language and ad length agnostic detection of televised advertisements

IN606027BActive Publication Date: 2026-10-14SILVEREDGE TECH PVT LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
IN201711019935
Authority / Receiving Office
IN · IN
Patent Type
Patents
Current Assignee / Owner
Filing Date
2017-06-07
Publication Date
2026-10-14
Estimated Expiration
2037-06-07

AI Technical Summary

Technical Problem

Current methods for detecting televised video advertisements are inefficient due to high error rates, computational inefficiency, and lack of language and platform agnosticism, leading to increased detection time and inaccurate differentiation between programs and ads.

Method used

A computer system using processors and memory to perform real-time detection of multi-lingual televised advertisements by selecting frames, extracting keypoints, deriving binary descriptors, and comparing them against a curated vocabulary of ad frames, facilitating language and channel agnostic detection.

Benefits of technology

The system achieves rapid, accurate, and precise detection of ads across multiple channels and platforms, reducing detection time and error rates, and is independent of ad length, language, and hardware.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The present disclosure provides a system for hardware, channel, language and ad length agnostic detection of multi-lingual televised advertisements. The detection is performed across live streams of media content of one or more broadcasted channels. The system selects a set of frames per second from a pre-defined set of frames. The system extracts a pre-defined number of keypoints from each selected frame and derivation of a pre-defined number of binary descriptors from the extracted keypoints. The system creates of a special pyramid of the binary descriptors and accessing a second vocabulary of binary descriptors. The system compares each spatially identifiable binary descriptor from the first vocabulary with spatially identifiable binary descriptors in clusters of the second vocabulary. The system progressively scores each selected frame and detects the first ad in the live streams of the media content. TO BE PUBLISHED WITH FIGURE 3
Need to check novelty before this filing date? Find Prior Art

Description

TECHNICAL FIELD

[0001] The present invention relates to the field of computer vision and machinelearning and, in particular, relates to a hardware, channel, language and ad lengthagnostic real time detection of televised video advertisements.BACKGROUND

[0002] With the proliferation and penetration of internet and televised medianetwork in recent years, the affordability and access of media networks hasconcurrently grown. With such penetration and public viewership, advertisershave shifted to the internet space and televised media networks for promotingtheir products and services. The demand for ad space has increased in thetelevised media networks concurrently with an increase in number of broadcastedchannels. These advertisers, publishers and content distributors need to closelymonitor airing of advertisements across multiple broadcasted television channelsand socials networks. The need for faster ad analytics across multiple channelsand platforms is witnessed with adoptions of competitive bidding betweenadvertisers for booking ad space.

[0003] A television broadcast essentially consists of videos of scheduledprograms and sponsored advertisements. Each advertisement video is generallyscheduled to run for 10 to 35 seconds approximately on multiple channels atdifferent or same time. The advertisements are provided by advertisers to run inbetween the scheduled broadcast of the program on each channel. Traditionally,these advertisements are either detected manually by assigning each user for eachchannel to records ads related data. In technological approaches, the use ofsupervised detection of key features of the ad in the stream of the broadcast isperformed. The approach focuses on detection of advertisements by extractingand analyzing digital audio fingerprints. The audio fingerprints areprobabilistically matched with similar fingerprints in a master database. Theprobabilistic match is analyzed for positive validation of the airing media as anad.

[0004] The present solutions have several disadvantages. T he present solutionsare inefficient for removing redundancy of false positives in ad detection. Theuse of audio fingerprints for detecting the ad increases the error rate with the useof similar audio tones and fingerprints in non-ad content. The time for detectionof the advertisement also increases with the usage of computationally inefficientand error prone techniques of supervised audio fingerprinting. The increase indetection time affects the synchronization and operations of cross platform adcampaigns. These solutions detect same ad with the different language ofcommunication on different native language channels as different ads. Thesesolutions are not language or completely platform agnostic. In addition, thesesolutions lack the precision and accuracy to differentiate programs fromadvertisements.

[0005] In light of the above stated discussion, there is a need for a method andsystem which overcomes the above stated disadvantages.SUMMARY

[0006] In an aspect, the present disclosure provides a computer system. Thecomputer system includes one or more processors and a memory coupled to theone or more processors. The memory causes the one or more processors toperform a method for hardware, channel, language and ad length agnostic realtime detection of multi-lingual televised advertisements. The real time detectionof the multi-lingual televised advertisements is performed across live streams ofmedia content of one or more broadcasted channels. The method includes a firststep of selecting a set of frames per second from a pre-defined set of frames. Theset of frames is selected from the pre-defined set of frames in each second of eachlive stream of the media content in each broadcasted channel of the one or morebroadcasted channels. The method includes another step of extracting a predefinednumber of keypoints from each selected frame of the media content. Thepre-defined number of keypoints is extracted on the basis of validation of at leastone of a scale invariance criterion, a rotation invariance criterion and a Harrisscore criterion. The method includes yet another step of deriving a pre-definednumber of binary descriptors from the pre-defined number of extracted keypoints.Each binary descriptor is characterized by a binary string with a length of 256bits. Further, the method includes yet another step of creating a special pyramidof the pre-defined number of binary descriptors to obtain a pre-defined number ofspatially identifiable binary descriptors. The special pyramid is created forobtaining a first vocabulary of binary descriptors of each selected frame. Themethod includes yet another step of accessing a second vocabulary of binarydescriptors for a curated comprehensive repository of ad frames. The curatedcomprehensive repository of the ad frames corresponds to a comprehensive set oftelevised advertisements. The second vocabulary includes a set of tree structuredclusters of binary descriptors. The method includes yet another step of comparingeach spatially identifiable binary descriptor from the first vocabulary of eachselected frame with a plurality of spatially identifiable binary descriptors in atleast one or more clusters of the second vocabulary of the binary descriptors. Themethod includes yet another step of progressively scoring each selected framefrom each live stream of the media content for validation of the selected frame asthe ad frame of a first ad. The first ad is progressively scored for each positivelyvalidated frame to obtain a progressive ad score. The progressive score iscalculated for the ad length independent detection of the first ad. The methodincludes yet another step of detecting the first ad in each live stream of the mediacontent of the one or more broadcasted channels when the progressive score forthe first ad is above a threshold score. The first ad is broadcasted by the one ormore broadcasted channels in one or more languages. The first ad is detectedindependent of the one or more languages of the first ad broadcasted in the one ormore broadcasted channels.BRIEF DESCRIPTION OF THE FIGURES

[0007] FIG. 1A illustrates a system for hardware, channel, language and adlength agnostic real time detection of multi-lingual televised advertisements, inaccordance with various embodiments of the present disclosure;

[0008] FIG. 1B illustrates an example of creating special pyramid for a selectedframe of the media content;

[0009] FIG. 1C illustrates an example of clustering binary descriptors intomillion clusters;

[0010] FIG. 1D illustrates an example of frame selections from the mediacontent for detecting ads and detection time thereof;

[0011] FIG. 2 illustrates a block diagram of an advertisement detection system,in accordance with various embodiments of the present disclosure; and

[0012] FIG. 3 illustrates a block diagram of a computing device, in accordancewith various embodiments of the present disclosure.DETAILED DESCRIPTION

[0013] FIG. 1A illustrates a system 100 for hardware, channel, language and adlength agnostic real time detection of multi-lingual televised advertisementsacross live streams of media content of one or more broadcasted channels, inaccordance with various embodiments of the present disclosure. The system 100performs real time supervised detection of televised video advertisements acrossthe live streams of the media content of the one or more broadcasted channels.The real time detection of televised video advertisements is significantlyinterrelated to the detection time for positive validation of airing content as atelevised video ad. The system 100 implements sophisticated statistical learningand mature computer vision techniques on video frames of the media content.The use of efficiently structured database management, feature extraction andanalysis provides robust and non-redundant detection of multi-lingual televisedvideo advertisements in the live streams of the media content. Moreover, thesystem 100 describes a network and computational infrastructure for coherent,concurrent and non-redundant reception, statistical analysis, recognition andsupervised detection of televised ads for one or more television channels. Theimplementation of computer vision, feature extraction and statistical learningtechniques on real time collected and pre-processed statistical models ofadvertisement videos facilitates the supervised detection of televised videoadvertisements. The media content in the live stream of the channel is selectivelyprocessed in units of frames and robust statistical learning models are developedover the selectively processed frames of the media content. The robust statisticallearning models are parametrically analyzed and prominent and prospectivefeatures are classified and compared for an improvement in time and error rate ofdetection of the televised video ads. The detection time is significantly reducedand ad analytics can be synchronized with social media or television adcampaigns.

[0014] The system 100 includes one or more channel media reception devices102 and an advertisement detection system 104. The above stated elements of thesystem 100 operate coherently and synchronously to detect televised ads in themedia content of the one or more broadcasted channels. Each channel mediareception device is a device for reception, authentication, decoding of the livestream of encoded media content in the form of channel feed of each broadcastedchannel. Each channel media reception device 102 is attached directly orindirectly to a receiving antenna or dish. The receiving antenna receives abroadcasted signal carrying one or more channel feeds. The one or more channelfeeds are encoded in a pre-defined format. In addition, the one or more channelfeeds have a set of characteristics. The set of characteristics include but may notbe limited to a frame rate, an audio sample rate, encoding scheme, transcodingscheme, bit rate and channel frequency band.

[0015] The broadcasted signal carrying the one or more channel feeds is initiallytransmitted from a transmission device. In an embodiment of the presentdisclosure, the broadcasted signal carrying the o ne or more channel feeds is amultiplexed MPEG-2 encoded signal having a constant bit rate. In anotherembodiment of the present disclosure, the broadcasted signal carrying the one ormore channel feeds is a multiplexed MPEG-2 encoded signal having a variable bitrate. In yet another embodiment of the present disclosure, the broadcasted signalcarrying the one or more channel feeds is any digital standard encoded signal.The bit rate is based on complexity of each frame in each of the one or morechannel feeds. The quality of the multiplexed MPEG-2 encoded signal will bereduced when the broadcasted signal is too complex to be coded at a constant bitrate.The bit rate of the variable bit-rate MPEG-2 streams is adjusted dynamicallyas less bandwidth is needed to encode the images with a given picture quality. Inaddition, the broadcasted signal is encrypted for a conditional access to aparticular subscriber. The encrypted broadcast signal is uniquely decoded by theone or more channel media reception devices 102 uniquely.

[0016] In an example, a digital TV signal is received on each channel mediareception device as a stream of MPEG-2 data. The MPEG-2 data has a transportstream. The transport stream has a data rate of 40 megabits / second for a cable orsatellite network. Each transport stream consists of a set of sub-streams. The setof sub-streams is defined as elementary streams. Each elementary streamincludes an MPEG-2 encoded audio, an MPEG-2 encoded video and dataencapsulated in an MPEG-2 stream. Moreover, each elementary stream includesa packet identifier (hereinafter "PID") that acts as a unique identifier forcorresponding elementary stream within the transport stream. The elementarystreams are split into packets in order to obtain a packetized elementary stream(hereinafter "PES").

[0017] In an embodiment of the present disclosure, the one or more channelmedia reception devices 102 are digital set top boxes. In another embodiment ofthe present disclosure, the one or more channel media reception devices 102 arehybrid set top boxes. In yet another embodiment of the present disclosure, theone or more channel media reception devices 102 are any standard broadcastsignal processing device. Each channel media reception device may receive thebroadcast signal from any broadcast signal medium. In an embodiment of thepresent disclosure, the broadcast signal medium is an Ethernet cable. In anotherembodiment of the present disclosure, the broadcast signal medium is a satellitedish. In yet another embodiment of the present disclosure, the broadcast signalmedium is a coaxial cable. In yet another embodiment of the present disclosure,the broadcast signal medium is a telephone line having DSL connection. In yetanother embodiment of the present disclosure, the broadcast signal medium is abroadband over power line (hereinafter "BPL"). In yet another embodiment ofthe present disclosure, the broadcast signal medium is wireless ordinary VHF orUHF bandwidth.

[0018] Each channel media reception device of the one or more channel mediareception devices 102 primarily includes a signal input port, an audio output port,a video output port, a de-multiplexer, a video decoder, an audio decoder and agraphics engine. The broadcast signal carrying the one or more channel feeds isreceived at the signal input port. The broadcast signal carrying the one or morechannel feeds is de-multiplexed by the de-multiplexer. The video decoderdecodes the encoded video and the audio decoder decodes the encoded audio.The video and audio corresponds to a channel selected in each channel mediareception device. Each channel media reception device of the one or morechannel media reception devices 102 carries the one or more channel feedsmultiplexed to form a single transporting stream.

[0019] The decoded audio and the decoded video for each live stream of eachbroadcasted channel are received at the audio output port and the video outputport. The decoded video is characterized by a set of features. The set of featuresinclude but may not be limited to a frame height, a frame width, a frame rate, avideo resolution, a bit rate and an encoding scheme. Further, the decoded videomay be of any display resolution. Examples of the display resolution include butmay not be limited to 144p, 240p, 360p, 480p, 720p, 1080p and 1080i. Here, pand i denotes progressive scan and interlace scan techniques.

[0020] The decoded video for each broadcasted channel is transferred to theadvertisement detection system 104. The decoded video corresponds to the livestream of media content for each broadcasted channel. The live streams of themedia content include ad streams and non-ad streams. The ad streams includeone or more ads. The non-ad streams may correspond to televised publishedcontent, channel related data, subscription related data, operator messages and thelike.

[0021] The media content is characterized by a pre-defined frame rate, a predefinednumber of frames and a pre-defined bit rate for a pre-defined interval ofbroadcast. The pre-defined frame rate is defined as a pre-defined set of framesstreamed per second in the live streams of the media content. The pre-defined setof frames per second is often referred as frame per second (FPS). Examples ofthe pre-defined set of frames per second include but may not be limited to 25FPS, 30 FPS, 20 FPS and 35 FPS.

[0022] The advertisement detection system 104 selects a set of frames persecond from the pre-defined set of frames. The pre-defined set of frames arestreamed in each second of each live stream of the media content for eachbroadcasted channel of the one or more broadcasted channels. Each selectedframe in the selected set of frames differs by a frame gap of at least 8 frames. Inan embodiment of the present disclosure, the selected set of frames per secondfrom the pre-defined set of frames is 3 per second. In another embodiment of thepresent disclosure, the selected set of frames per second from the pre-defined setof frames is more or less than 3 per second. For example, the advertisementdetection system 104 may be configured to select every 8th frame from 25 framesper second of the media content. The pre-defined set of frames may be every 8th,16th and 24th frame per 25 frames in each second of the media content.

[0023] The advertisement detection system 104 performs histogramnormalization on each selected frame from the selected set of frames of the mediacontent broadcasted on each broadcasted channel. The advertisement detectionsystem 104 sets each pixel value in each frame of the media content to a standardpixel value range set. Further, the advertisement detection system 104 scales eachframe of the media content to a pre-defined scale. In an embodiment of thepresent disclosure, the pre-defined scale of each frame is 640 by 480. Eachselected frame of the set of selected frame per second is characterized by a predefinednumber of pixels. Each pixel is fundamentally a combination of Red,Green and Blue colors (hereinafter "RGB"). Each color of the pixel (RGB) hasthe pre-defined value in a range of 0-255. The advertisement detection system104 gray-scales each selected frame of the selected set of frames. Theimplementation of gray-scaling and histogram normalization on each selectedframe removes any hardware dependency of the selected set of frames in themedia content across the live streams of the one or more broadcasted channels.Moreover, the removal of the hardware dependency facilitates in hardwareagnostic supervised detection of the televised video advertisements across the livestreams of the media content of the one or more broadcasted channels.

[0024] The advertisement detection system 104 extracts a pre-defined number ofkeypoints from each selected frame of the selected set of frames per second of themedia content. Each extracted keypoint is characterized by a spatial position ineach selected frame. In general, each keypoint in any gray-scaled framecorresponds to a pixel with enough neighborhood pixels in a different gray areawith the pixel. Each keypoint is generally selected by choosing an arbitrary pixelpoint as center to form a circular area and assigning the circular area as the pixelpoint's neighborhood. The circular area of the selected keypoint is analyzed forsatisfying the condition:Where IX is the intensity of gray-scaled neighborhood pixels and IP is theintensity of a selected feature pixel.

[0025] The advertisement detection system 104 uses Harris detector to select thepre-defined number of keypoints and sorts the selected pre-defined set ofkeypoints. The pre-defined number of keypoints is extracted based on validationof at least one of scale invariance criterion, rotation invariance criterion, Harrisscore criterion and translational invariance criterion. The extracted pre-definednumber of keypoints is invariant to at least one of rotation, translation and scalingof objects in selected frame in the set of frames of the media content. In anexemplary embodiment of the present disclosure, the pre-defined number ofkeypoints for each selected frame in the selected set of frames is 700. In anotherembodiment of the present disclosure, the pre-defined number of keypoints foreach selected frame in the selected set of frames is more than 700. In yet anotherembodiment of the present disclosure, the pre-defined number of keypoints foreach selected frame in the selected set of frames is less than 700. The number ofselections depends on a scale of detection and limited computational capacity ofthe advertisement detection system 104.

[0026] The advertisement detection system 104 calculates direction in corners ofeach keypoint. The directions in the form of intensity vectors are calculated usinguses intensity centroid technique. The advertisement detection system 104|IX – IP| > ITderives a pre-defined number of binary descriptors from the pre-defined numberof extracted keypoints. Each binary descriptor is characterized by a binary stringwith a length of 256 bits. Each 256 bit of binary descriptor uniquely identifies afeature point in the selected frame of the media content. Each binary descriptor isderived after adaptive implementation of rotational, translational and scaleinvariance techniques. The number of binary descriptors depends on the numberof extracted keypoints. In an exemplary embodiment of the present disclosure,the pre-defined number of binary descriptors is 700. In another embodiment ofthe present disclosure, the pre-defined number of binary descriptors is more than700. In yet another embodiment of the present disclosure, the pre-defined numberof binary descriptors is less than 700.

[0027] The advertisement detection system 104 is a computing server withprovisions for volatile memory as random access memory (RAM) and permanentnon-volatile storage spaces. The advertisement detection system 104 maintainsone or more databases in the volatile memory and permanent non-volatile storagespace. The one or more databases in the advertisement detection system 104includes a first vocabulary database 104a, a second vocabulary database 104b, anad frame repository 104c and an ad detection database 104d. The advertisementdetection system 104 uses the one or more databases for storing the descriptordata and other statistical and detection data.

[0028] The advertisement detection system 104 stores the derived descriptordata in the first vocabulary database 104a. The advertisement detection system104 assigns a weight value to each binary descriptor of the pre-defined number ofbinary descriptors in the first vocabulary database 104a. The assignment of theweight values to the binary descriptors is based on a term frequency and aninverse document frequency of the descriptors in the first vocabulary database104a. For example, the weight value for a binary descriptor (B) with a higherterm frequency will be less than the weight value for a binary descriptor (C) witha lower term frequency. The weight values of the binary descriptors are L1normalized to lie in a range of 0 and 1. The normalized weight values for thebinary descriptors are characterized by an arithmetic sum of 1.

[0029] Mathematically, if Wi denotes the weight values of N descriptors, thenthe L1 normalization of the Wi weights is defines as:

[0030] The normalization of the weight values of the pre-defined number of thebinary descriptors is performed to simplify calculation parameters and reducestatistical sparse distribution of the weight values. For example, an assignedweight value W1 is 30 for a specific term frequency and another weight value W2is 150 for another term frequency. The L1 normalization of W1 and W2 will be30 / (30+150) and 150 / (30+150), i.e. 0.167 and 0.833 respectively. The L1normalized weights of the pre-defined number of binary descriptors are stored inthe first vocabulary database 104a of the advertisement detection system 104.

[0031] The spatial information of the extracted keypoint and associated binarydescriptor is lost with the derivation of the binary descriptors for the extractedkeypoints from each selected frame of the media content. The spatial informationis revalidated and identified by creating a special pyramid of binary descriptors(as shown in FIG. 1B). The advertisement detection system 104 creates thespecial pyramid of the pre-defined number of derived binary descriptors to obtaina pre-defined number of spatially identifiable binary descriptors. The specialpyramid is a database of binary descriptors with three stages of each selectedframe. The three stages are labeled for illustration in FIG. 1B. The three stagesare stage 1, stage 2 and stage 3 respectively. The three stages are segregated into(equation)nine discrete levels for the same selected frame. The nine discrete levels arelabeled in FIG. 1B as L1, L2, L3, L4, L5, L6, L7, L8 and L9 respectively. Eachlevel segregates the keypoints and descriptors according to a specific slicedregion in the selected frame. The first stage includes the entire selected frame andthe binary descriptors as sparsely distributed over the entire frame in the firstlevel (L1). The spatial information is uncertain in L1. The second stage includesa second level (L2) and a third level (L3) of the binary descriptors as slices of theselected frame into two equal rectangular halves. The second level (L2) maps to afirst fraction of the binary descriptors and the third level (L3) maps to remainingbinary descriptors. The third stage includes six levels (L4, L5, L6, L7, L8 and L9)of the binary descriptors. The six levels (L4, L5, L6, L7, L8 and L9) aredistributed over six slices of the selected frame. Each slice maps a specificfraction of the binary descriptors. Each binary descriptor can be mapped spatiallyin each stage and subsequent level of the special pyramid. The third stage is mostsignificant for spatially identifying each binary descriptor in the pre-definednumber of the binary descriptors. The number of spatially identifiable binarydescriptors is thrice the number of the binary descriptors for each selected frame.In an embodiment of the present disclosure, the pre-defined number of spatiallyidentifiable binary descriptors is 2100 when the pre-defined number of binarydescriptors is 700. The pre-defined number of spatially identifiable binarydescriptors are updated, sorted and stored as records in the first vocabularydatabase 104a (as shown in FIG. 1B). Also, each spatially identifiable binarydescriptor is tagged with a level identifier in the first vocabulary database 104a.Each weight value for each binary descriptor in the special pyramid of the pre25defined number of spatially identifiable binary descriptors is again L1normalized. Each weight value in the special pyramid is L1 normalized such thatthe arithmetic sum of the weight values is 1. The advertisement detection system104 finally obtains a first vocabulary of binary descriptors for the pre-definednumber of spatially identifiable binary descriptors of each selected frame in thefirst vocabulary database 104a.

[0032] The detection of a first ad in the live stream of the media content is basedon implementation of supervised detection techniques. The advertisementdetection system 104 accesses a second vocabulary of binary descriptors in thesecond vocabulary database 104b. The second vocabulary of binary descriptorscorresponds to a curated comprehensive repository of ad frames from acomprehensive set of televised advertisements. The curated comprehensiverepository of ad frames from the comprehensive set of televised advertisements isstored in the ad frame repository 104c of the advertisement detection system 104.The second vocabulary of the binary descriptors is created prior to real timedetection of the live stream of the media content. The second vocabulary ofbinary descriptors is elementally stored in the second vocabulary database 104bas spatially identifiable records with tagged levels of the special pyramid andnormalized weight values. The advertisement detection system 104 creates thesecond vocabulary of binary descriptors from the comprehensive repository of adframes. The advertisement detection system 104 extracts the pre-defined numberof keypoints and the pre-defined number of binary descriptors from each frame inthe repository of ad frames. The advertisement detection system 104 creates thespecial pyramid of the binary descriptors for each ad frame to obtain the predefinednumber of spatially identifiable binary descriptors.

[0033] The advertisement detection system 104 clusters the binary descriptorsinto a first set of clusters (as shown in FIG. 1C). The binary descriptors areclustered into the first set of clusters. The first set of clusters is created based onan evaluation of a minimum hamming distance between each binary descriptor inthe pre-defined number of spatially identifiable binary descriptors. Each clusterin the first set of clusters is associated with a centroid. The centroid of eachcluster is a middle of each cluster. The centroid is a vector which includes anumber for each variable in each cluster of the first set of clusters. Each numberis a mean of the variables in each cluster of the first set of clusters. Each centroidis measure of location of each cluster in the set of clusters. In an embodiment ofthe present disclosure, the first set of clusters includes 10 centroids associatedwith 10 clusters (C1-C10). In another embodiment of the present disclosure, thefirst set of clusters includes more or less than 10 centroids associated with moreor less than 10 clusters. Each cluster in the first set of clusters includes a fractionof the pre-defined number of spatially identifiable binary descriptors. Further, theadvertisement detection system 104 iteratively clusters the binary descriptors ineach cluster of the first set of clusters and each subsequent cluster for a predeterminediteration level to obtain a set of tree structured clusters. The predeterminediteration level for iteratively clustering the derived binary descriptorsobtained from the repository of ad frames is 6 (as shown in FIG. 1C). Eachcluster in the set of tree structured clusters is associated with a centroid. In anembodiment of the present disclosure, the set of tree structured clusters includes106 centroids in association with 106 clusters (C1 (105) - C 10(105)) when the predeterminediteration level is 6. Each binary descriptor in each cluster of the set oftree structured clusters is optimally clustered for a lowest hamming distanceamong the fraction of the binary descriptors.

[0034] The advertisement detection system 104 assigns the weight value to eachclustered binary descriptor in the set of tree structured clusters. Each weightvalue is evaluated based on optimal adjustment of the term frequency and theinverse document frequency. Further, each weight value is L1 normalized suchthat the evaluated arithmetic sum of weight values is 1. The binary descriptorsand associated L1 normalized weight values in the set of tree structured clustersare optimally stored in a distributed chain of structured and sorted tables as asecond vocabulary of binary descriptors. The second vocabulary of the binarydescriptors is stored in the second vocabulary database 104b. Furthermore, thedistributed chain of tables includes one or more descriptor tables and one or morelook up tables. The one or more look up tables includes records pointing eachclusters and associated descriptors in each clusters of the set of tree structuredclusters. The second vocabulary includes the set of tree structured clusters ofbinary descriptors as leaf nodes. Each leaf node is associated with normalizedweight values. The second vocabulary of the binary descriptors is characterizedas an n-ary tree data structure with the leaf nodes.

[0035] In an exemplary example, a large collection of televised videoadvertisements, say 2000 are stored in the permanent non-volatile storage of theadvertisement detection system 104. Considering each ad to have an averagelength of 15 seconds, the advertisement detection system 104 may create thecomprehensive repository of ad frames from the 2000 ads. If each ad is run at 25FPS for 15 seconds, then total number of frames for the ad will be 375 frames.The total number of frames for 2000 ads will be roughly 750000 frames for 25FPS. Let us suppose that the advertisement detection system 104 extracts roughly250000 prominent frames to create the comprehensive 250000 strong repositoryof ad frames.

[0036] Referring to the aforementioned example, the advertisement detectionsystem 104 may extract, say 700 keypoints for each ad frame of the 250000 adframes in the repository of ad frames. The total number of keypoints for 250000ad frames is 175 million keypoints. The advertisement detection system 104derives 175 million binary descriptors against the 175 million keypoints. Theadvertisement detection system 104 performs iterative clustering on the 175million binary descriptors. In first iteration, the first set of clusters is obtained.Each set in the first set of clusters includes a fraction of 175 million descriptors.Let us assume that the 1st cluster, 2nd cluster, 3rd cluster, 4th cluster, 5th cluster, 6thcluster, 7th cluster, 8th cluster, 9th cluster and 10th cluster includes 20, 22, 30, 25,23, 22, 8, 4, 12, 7 million binary descriptors respectively. Each cluster is obtainedon the basis of minimum inter-binary descriptor hamming distance. Let us takesix 8 bit binary numbers as an example for six 256 bit binary descriptors. The 8bit descriptors are 01101010, 01011100 and 10101010 respectively. Thehamming distance between the first descriptor and the second descriptor, thesecond descriptor and the third descriptor and the third and fourth descriptor is 4,6 and 2. The advertisement detection system 104 segregates and clusters thebinary descriptors with hamming distance of 4, 6 and 2 respectively.

[0037] In second iteration, each cluster in the first set of cluster is againclustered into 10 different clusters. The total number of clusters obtained aftersecond iteration is 102. In third iteration, each cluster in 102 clusters is againclustered to obtain 103 clusters. In fourth iteration and fifth iteration, each clusterin the 103 and 104 clusters is again clustered. The iterative clustering is done tillsixth iteration and a total of 106 clusters are obtained. Each cluster containingapproximately 1 million binary descriptors. 106 clusters are stored and structuredas the n-ary tree data structure with each cluster at each leaf node of the n-arytree. The advertisement detection system 104 creates the special pyramid of thebinary descriptors in the 106 clusters for validating spatial information of thekeypoints. The tree structured clusters are stored as the second vocabulary in thesecond vocabulary database 104b.

[0038] The advertisement detection system 104 compares each spatiallyidentifiable binary descriptor in the first vocabulary with spatially identifiablebinary descriptors in at least one or more clusters of the set of tree structuredclusters. The set of tree structured clusters correspond to the second vocabularyof the binary descriptors of the repository of the ad frames. The advertisementdetection system 104 searches for a cluster in the set of tree structured clustersthat includes binary descriptors with minimum hamming distances with the binarydescriptor in the first vocabulary of the selected frame in the live stream of themedia content. Each binary descriptor from the first vocabulary of the binarydescriptors for the selected frame of the media content is matched with the binarydescriptor in the searched cluster for the minimum hamming distance. The binarydescriptors are matched frame by frame. The advertisement detection system 104calculates a Bhattacharya distance between weight values of each matched pair ofthe binary descriptors. The Bhattacharya distance is defined as square root of aproduct of a first weight value and a second weight value. The advertisementdetection system 104 sums the calculated Bhattacharya distance for each matcheddescriptor in the first vocabulary of the binary descriptors for the selected frameof the media content to obtain a summed feature value. The summed featurevalues lies in a normalized range of 0 to 1.

[0039] If Wi is the weight value for an ith binary descriptor in the firstvocabulary and W j is the weight value of the jth binary descriptor in the set oftree structured clusters, then the feature value (FV) is calculated as:Where i and j denote an index of the binary descriptor in the first vocabulary andthe second vocabulary of binary descriptors respectively and N denotes thenumber of the binary descriptors in the first vocabulary of the binary descriptorsand where:

[0040] For example, let us assume that the 700 binary descriptors in the firstvocabulary of the selected frame are matched with binary descriptors in one of theclusters of the second vocabulary of the binary descriptors. Each pair of matcheddescriptors in the first vocabulary of the binary descriptors and the secondvocabulary of the binary descriptors is characterized by the weight values. Theadvertisement detection system 104 calculates the Bhattacharya distance between(equation)the weight values of the matched pair by taking the square root of the product ofthe weight values. The advertisement detection system 104 calculates the featurevalue (FV) by summing the Bhattacharya distance between the matched pair ofthe binary descriptors. The summed feature value lies in a range of 0 to 1.

[0041] The advertisement detection system 104 progressively scores eachselected frame across each live stream of the media content for validation of theselected frame as ad frame of the first ad. The first ad is assumed to be present inthe live stream of the one or more broadcasted channels. Moreover, the first ad isbroadcasted by the one or more broadcasted channels in one or more languages.The one or more languages of the first ad are present as an audio stream in thefirst ad. The advertisement detection system 104 adopts video processingtechniques for supervised detection of the first ad and avoids the processing oflanguage dependent audio stream of the first ad. The implementation ofsupervised video processing techniques facilitates the one or more languagesindependent detection of the first ad across the live streams of the one or morebroadcasted channels.

[0042] The detection of the first ad in the live streams of the media content isbased on progressive evaluation of the summed feature values for each selectedframe in the set of frames of the media content. Each selected frame for the firstad in the live streams of the media content is progressively scored in at least oneor more steps. The ad may be positively evaluated at any of the steps in the oneor more steps. The one or more steps include a first step of comparison test. Theadvertisement detection system 104 compares the summed feature value for eachselected frame in the set of frames per second of the media content with a firstthreshold value in the comparison test. The summed feature value is comparedfor progressive validation of the selected frame as the ad frame. In anembodiment of the present disclosure, the first threshold value for validation ofthe selected frame as ad frame is 0.02. The advertisement detection system 104evaluates a ratio test between selected frames in the selected set of frames of themedia content. The ratio test is evaluated for determination of a degree ofdifference in consecutively selected frames in the set of frames per second of themedia content. The ratio test is evaluated by dividing summed feature value for asecond frame by corresponding summed feature value for a first frame in theselected set of frames. The advertisement detection system 104 rewards a firstscore for the first ad of the one or more ads in the live streams of the mediacontent. The first ad is streamed in real time in the live stream of the mediacontent. The first ad is rewarded by assigning a first ad score for the positivevalidation of the evaluated ratio. The first score is assigned to the first ad scorewhen the evaluated ratio is less than a second threshold value. In an embodimentof the present disclosure, the first ad may be assigned a first ad score of 1.5 whenthe ratio of feature value for the second frame to that of the first frame is less thanthe second threshold value of 0.7. In another embodiment of the presentdisclosure, the first ad may be assigned a different first ad score when the ratio offeature value for the second frame to that of the first frame is less than a differentsecond threshold value.

[0043] For example, a first frame (A) may have a feature value of 0.11 and asecond frame (B) may have a feature value of 0.09. The ratio of the feature valuefor the second frame (B) to that of the first frame (A) is 0.07 / 0.11 or 0.636. Theratio test evaluates to positive evaluation as the evaluated ratio is less than 0.7.The first ad is assigned a first ad score of 1.5 for the first frame.

[0044] The advertisement detection system 104 penalizes a second ad score for asecond ad of the one or more ads of the media content by deducting a secondscore from the assigned first second ad score for the second ad. The second ad isa past ad streamed before the first ad. As the first ad is progressively detected inthe live stream with progressive additions of score, the second ad is progressivelydevaluated with progressive deductions of the second score from the second adscore of the second ad. In an embodiment of the present disclosure, the secondscore as penalty to the second ad score of the second ad is 0.5. In anotherembodiment of the present disclosure, the second score as penalty to the secondad score of the second ad is more or less than 0.5.

[0045] The advertisement detection system 104 performs an equality test forconsecutive frames in the selected set of frames of the media content. Theequality test is based on equality criterion between the feature value of the firstframe and the feature value of the second frame in the selected set of frames. Theadvertisement detection system 104 rewards the first ad score for the first ad ofthe one or more ads in the live stream of the media content with additions of athird score to the first ad score. The first ad score is rewarded with the third scorewhen the feature value of the first frame is equal to the feature value of the secondframe in the selected set of frames. In an embodiment of the present disclosure,the third score as reward to the first ad score of the first ad is 0.5. In anotherembodiment of the present disclosure, the third score as reward to the first adscore of the first ad is more or less than 0.5.

[0046] The advertisement detection system 104 performs a vicinity test forascertaining neighboring frames as the ad frames of the first ad. The vicinity testis needed to robustly detect video ads of different duration. The vicinity testmakes the detection length agnostic. The advertisement detection system 104rewards the first ad score for the first ad of the one or more ads in the live streamof the media content with addition of a fourth score to the first ad score. Thefourth score is rewarded on vicinity criterion and the vicinity criterion is based onsuccessive positive validation of the first score for the first frame and the secondframe of the set of frames. In an embodiment of the present disclosure, the fourthscore as reward to the first ad score of the first ad is 0.5. In another embodimentof the present disclosure, the fourth score as reward to the first ad score of thefirst ad is more or less than 0.5. The advertisement detection system 104calculates a progressive ad score for the first ad and the second ad. Theprogressive score is calculated on the basis of at least one of progressive additionand subtraction of the second score, the third score and the fourth score to the firstad score. The first ad is progressively scored for each positively validated frameto obtain the progressive ad score. The progressive s core is calculated for the adlength independent detection of the first ad. For example, the first ad may bebroadcasted in a first live stream of a first broadcasted channel for 10 seconds andin a second live stream of a second broadcasted channel for 15 seconds. The 10seconds and 15 seconds version of the first ad differ by few frames. Theadvertisement detection system 104 uses vicinity test to evaluate whether theselectively processed frames in the vicinity of the positively tested ad frames arerelated to the same first ad. The advertisement detection system 104 facilitatesthe ad length agnostic detection of the first ad across the one or more broadcastchannels.

[0047] For example, let us suppose a first frame (F1) of the first ad is positivelyevaluated. The first ad may have an ad score of 5.5. The subsequent frames (F1and F2) are evaluated on the basis of the vicinity criterion. The vicinity criterionis positively validated when consecutive frames in the neighborhood of the firstframe (F1) are positively validated. The ad score is increased by 0.5 for eachpositive validation. The advertisement detection system 104 calculates theprogressive score for the first ad as the sum of 3 and 0.5 as 6. The progressivescore may be analyzed for a specific duration of the live stream of the mediacontent. If the evaluated score is greater than or equal to a specified thresholdscore of, say 6, then the ad is said to be positively detected.

[0048] The advertisement detection system 104 detects the first ad in the livestreams of the media content when the progressive score for the first ad is above athreshold score. In an embodiment of the present disclosure, the threshold scorefor the positive detection of the first ad is 6. In another embodiment of thepresent disclosure, the threshold score for the positive detection of the first ad inthe live streams of the media content is more than 6. In yet another embodimentof the present disclosure, the threshold score for the positive detection of the firstad is greater than 1.5. In yet another embodiment of the present disclosure, thethreshold score for the positive detection of the first ad in the live streams of themedia content is based on any suitable range.

[0049] The advertisement detection system 104 detects the first ad in the livestreams of the media content in a detection period (DP) (as shown in FIG. 1D).The detection period (DP) is evaluated as equivalent of positive frame detectionfor at least two consecutively selected frames per second. For example, let ussuppose that for selection of 3 frames from 25 frames per second, theadvertisement detection system 104 positively evaluates the first two or the lasttwo frames as the ad frames. Moreover, the time for reception of each frame ofthe 3 selected frames per second is 0.33 x 2, i.e. 0.66 seconds and therefore, thedetection period (DP) for detection of each ad in the live streams of the mediacontent is equivalent to reception of at least two frames of each ad. The detectionperiod lies in a range of 0.6 seconds to 1 second.

[0050] Referring to FIG. 1D, the pre-defined set of frames are illustrativelylabeled with an index from 1 to 25 for a second of the media content from thebroadcasted channel. The advertisement detection system 104 selects 3 framesfrom every 8th position of 1 second of the media content. The three selectedframes are labeled 8, 16 and 24 for 8th frame, 16th frame and 24th frame. Thedetection period (DP) for positive detection of the first ad in the live streams ofthe media content is equivalent to selection time for at least two frames of themedia content. The illustrated delay between selections of every consecutivelyselected frame is illustrated as 0.33 seconds. The detection period (DP) is at leasttwice of the selection time for the at least two frames. The detection period (DP)is illustrated as 0.66 seconds.

[0051] The advertisement detection system 104 may ascertain a length of thefirst ad when the first ad score for the first ad gets penalized for a first time. Theadvertisement detection system 104 stores the detected ad information andassociated metadata in the ad detection database 104d. The ad detection database104d is updated with positive detection of the first a d detection in the live streamsof the media content in real time.

[0052] It may be noted that in FIG. 1A, FIG. 1B, FIG. 1C and FIG. 1D, theadvertisement detection system 104 detects the first ad in the media content of theone or more broadcasted channels; however, those skilled in the art wouldappreciate that the advertisement detection system 104 detects different ads in themedia content of the one or more broadcasted channels in the real time.

[0053] FIG. 2 illustrates a block diagram 200 of the advertisement detectionsystem 104, in accordance with various embodiments of the present disclosure.The block diagram 200 describes the advertisement detection system 104. Theadvertisement detection system 104 hardware, channel, language and ad lengthagnostic real time detection of multi-lingual televised advertisements across thelive streams of media content of one or more broadcasted channels. The livestream of the media content for each broadcasted channel is received andprocessed through implementation of computer vision, feature extraction;statistical learning techniques. The supervised detection of televised videoadvertisements is facilitated by real time collection of features and developmentof statistical models. The media content in the live stream of each broadcastedchannel is selectively processed frame by frame and robust statistical learningmodels are developed over the selectively processed frames of the media content.The robust statistical learning models are parametrically analyzed and prominentand prospective features are classified and compared for an improvement in timeand error rate of detection of the televised video ads.

[0054] The block diagram 200 of the advertisement detection system 104includes a reception module 202, a selection module 204, an extraction module206, a deriving module 208 and an assignment module 210. Moreover, the blockdiagram 200 includes a creation module 212, a normalization module 214, anobtaining module 216, an access module 218, a comparison module 220, ascoring module 222 and a detection module 224.

[0055] The reception module 202 receives the live stream of the media contentfrom the broadcasted channel. The live stream of the media content includes of alive ad stream and a non-ad stream. The live ad stream includes the one or moreads (as discussed in detailed description of FIG. 1A). Further, the selectionmodule 204 selects the set of frames per second from the pre-defined set offrames in each second of the live stream of the media content. Each selectedframe in the selected set of frames differs by a frame gap of at least 8 frames. Inan example, the set of frames is 3 per second from a 25 frames per second of thelive stream of the media content (as discussed in the detailed description of FIG.1A).

[0056] The extraction module 206 extracts the pre-defined number of keypointsfrom each selected frame. Each extracted keypoint is characterized by a spatialposition in selected frame. The pre-defined number of key points is extractedbased on validation of at least one of scale invariance criterion, a rotationinvariance criterion and Harris score criterion. In an embodiment of the presentdisclosure, the pre-defined number of keypoints is 700 (as discussed in detaileddescription of FIG. 1A). The deriving module 208 derives the pre-definednumber of binary descriptors from pre-defined number of extracted keypoints.Each binary descriptor is characterized by a binary string with a length of 256 bits(as discussed in detailed description of FIG. 1A). The assignment module 210assigns the weight value to each binary descriptor of the pre-defined number ofbinary descriptors. The weight value for each binary descriptor is L1 normalized.Each normalized weight value for each binary descriptor is characterized by anarithmetic sum of 1 (as discussed in detailed description of FIG. 1A).

[0057] The creation module 212 creates the special pyramid of the pre-definednumber of derived descriptors to obtain the pre-defined number of spatiallyidentifiable binary descriptors. The special pyramid of the pre-defined number ofbinary descriptors includes one or more copies of derived binary descriptors.Each binary descriptor in the special pyramid is tagged with an identifier forspatially segregating and validating position of each binary descriptor in theselected frame (as discussed in the detailed description of FIG. 1A). Thenormalization module 214 normalizes each weight value for each binarydescriptor in the special pyramid of the pre-defined number of spatiallyidentifiable binary descriptors. Each weight value in the special pyramid is L1normalized such that the arithmetic sum of the weight values is 1 (as discussed inthe detailed description of FIG. 1A).

[0058] The obtaining module 216 obtains the first vocabulary of binarydescriptors of each selected frame. The first vocabulary of binary descriptorscorresponds to the pre-defined number of spatially identifiable binary descriptors(as discussed in the detailed description of FIG. 1A). The access module 218accesses the second vocabulary of binary descriptors from the comprehensive setof televised advertisements. The second vocabulary of binary descriptorscorresponds to the curated comprehensive repository of the ad frames. Thesecond vocabulary of the binary descriptors is characterized by an n-ary tree datastructure with leaf nodes. The second vocabulary includes the set of treestructured clusters of binary descriptors as the leaf nodes and associated withweight values. The second vocabulary of the binary descriptors is created fromextraction of the pre-defined number of keypoints and the binary descriptors fromeach frame of the repository of ad frames. The special pyramid of the descriptorsis created for each ad frame to obtain the pre-defined number of spatiallyidentifiable binary descriptors. The binary descriptors are clustered into the firstset of clusters. The binary descriptors are clustered into the first set of clustersbased on the evaluation of minimum hamming distance between each binarydescriptor. The binary descriptors in each cluster of the first set of clusters areiteratively clustered. Each subsequent cluster is clustered for the pre-determinediteration level to obtain the set of tree structured clusters. In an embodiment ofthe present disclosure, the pre-determined iteration level is 6. The weight value,the term frequency and the inverse document frequency are assigned to eachclustered binary descriptor and each weight value is normalized using L1normalization such that the evaluated arithmetic sum of weight values is 1 (asdiscussed in the detailed description of FIG. 1A).

[0059] The comparison module 220 compares each spatially identifiable binarydescriptor from the first vocabulary of each selected frame with the plurality ofspatially identifiable binary descriptors in the at least one or more clusters. Theone or more clusters are searched from the set of tree structured clusters. The setof tree structured clusters corresponds to the second vocabulary of the binarydescriptors of the repository of the ad frames. The binary descriptors in the firstvocabulary are compared with the second vocabulary (as described in the detaileddescription of FIG. 1A).

[0060] The scoring module 222 progressively scores each selected frame fromthe live stream of media content for the validation of the selected frame as the adframe of the first ad. The first ad is progressively scored for each positivelyvalidated frame to obtain the progressive ad score (as described in the detaileddescription of FIG. 1A). The detection module 224 detects the first ad in the livestream of the media content when the progressive score for the first ad is above athreshold score. In an example, the threshold score for positive detection of thefirst ad in the live stream of the media content is 6 (as described in the detaileddescription of FIG. 1A).

[0061] FIG. 3 illustrates a block diagram of a computing device 300, inaccordance with various embodiments of the present disclosure. The computingdevice 300 includes a bus 302 that directly or indirectly couples the followingdevices: memory 304, one or more processors 306, one or more presentationcomponents 308, one or more input / output (I / O) ports 310, one or moreinput / output components 312, and an illustrative power supply 314. The bus 302represents what may be one or more busses (such as an address bus, data bus, orcombination thereof). Although the various blocks of FIG. 3 are shown withlines for the sake of clarity, in reality, delineating various components is not soclear, and metaphorically, the lines would more accurately be grey and fuzzy. Forexample, one may consider a presentation component such as a display device tobe an I / O component. Also, processors have memory. The inventors recognizethat such is the nature of the art, and reiterate that the diagram of FIG. 3 is merelyillustrative of an exemplary computing device 300 that can be used in connectionwith one or more embodiments of the present invention. Distinction is not madebetween such categories as "workstation," "server," "laptop," "hand-held device,"etc., as all are contemplated within the scope of FIG. 3 and reference to"computing device."

[0062] The computing device 300 typically includes a variety of computer20readable media. The computer-readable media can be any available media thatcan be accessed by the computing device 300 and includes both volatile andnonvolatile media, removable and non-removable media. By way of example,and not limitation, the computer-readable media may comprise computer storagemedia and communication media. The computer storage media includes volatileand nonvolatile, removable and non-removable media implemented in anymethod or technology for storage of information such as computer-readableinstructions, data structures, program modules or other data. The computerstorage media includes, but is not limited to, RAM, ROM, EEPROM, flashmemory or other memory technology, CD-ROM, digital versatile disks (DVD) orother optical disk storage, magnetic cassettes, magnetic tape, magnetic diskstorage or other magnetic storage devices, or any other medium which can beused to store the desired information and which can be accessed by the computingdevice 300. The communication media typically embodies computer-readableinstructions, data structures, program modules or other data in a modulated datasignal such as a carrier wave or other transport mechanism and includes anyinformation delivery media. The term "modulated data signal" means a signalthat has one or more of its characteristics set or changed in such a manner as toencode information in the signal. By way of example, and not limitation,communication media includes wired media such as a wired network or directwiredconnection, and wireless media such as acoustic, RF, infrared and otherwireless media. Combinations of any of the above should also be included withinthe scope of computer-readable media.

[0063] Memory 304 includes computer-storage media in the form of volatileand / or nonvolatile memory. The memory 304 may be removable, nonremovable,or a combination thereof. Exemplary hardware devices include solidstatememory, hard drives, optical-disc drives, etc. The computing device 300includes one or more processors that read data from various entities such asmemory 304 or I / O components 312. The one or more presentation components308 present data indications to a user or other device. Exemplary presentationcomponents include a display device, speaker, printing component, vibratingcomponent, etc. The one or more I / O ports 310 allow the computing device 300to be logically coupled to other devices including the one or more I / Ocomponents 312, some of which may be built in. Illustrative components includea microphone, joystick, game pad, satellite dish, scanner, printer, wireless device,etc.

[0064] The present disclosure has numerous disadvantages over the prior art.The present disclosure provides ad length independent, hardware or platformindependent, language independent and channel independent detection ofadvertisements. The present disclosure solves the problem of larger detectiontime of at least 7 seconds for a 15 second to 30 second ad. The present disclosureprovides a statistical and parametric method for robust detection of televised adsacross many broadcast channels in a very short duration of 0.6 seconds to 1second against earlier detection time of at least 7 seconds. The detection isprogressive with a very low error rate in false detection and computationally lightweighted. The video ads are detected within reception of first two frames of thelive stream. The detection is agnostic to platform, broadcast channel, receivinghardware, ad length and noise. The detection is supported by highly efficientdatabase management, supervised machine learning and advanced customcomputer vision techniques. The binary descriptors are selected after efficientconsideration of feature redundancy, rotational invariance, translationalinvariance and Harris score. The use of validation criterion on the basis ofprogressively awarded and penalized ad scores with each positive and negativevalidation provides a robust means to ascertain whether the detection media is anad video.

Claims

1. A computer system comprising: one or more processors; and a memory coupled to the one or more processors, t he memory for storing instructions which, when executed by the one or more processors, cause the one or more processors to perform a method for hardware, channel, language and ad length agnostic real time detection of multi-lingual televised advertisements across live streams of media content of one or more broadcasted channels, the method comprising: selecting, at an advertisement detection system, a set of frames per second from a pre-defined set of frames in each second of each live stream of the media content in each broadcasted channel of the one or more broadcasted channels; extracting, at the advertisement detection system, a pre-defined number of keypoints from each selected frame of the media content, the pre-defined number of keypoints being extracted based on validation of at least one of a scale invariance criterion, a rotation invariance criterion and a Harris score criterion; deriving, at the advertisement detection system, a pre-defined number of binary descriptors from corresponding pre-defined number of extracted keypoints, each binary descriptor being characterized by a binary string with a length of 256 bits; creating, at the advertisement detection system, a special pyramid of the pre-defined number of binary descriptors to obtain a pre-defined number of spatially identifiable binary descriptors, the special pyramid being created for obtaining a first vocabulary of binary descriptors corresponding to the predefined number of spatially identifiable binary descriptors of each selected frame; accessing, at the advertisement detection system, a second vocabulary of binary descriptors corresponding to a curated comprehensive repository of ad frames from a comprehensive set of televised advertisements, wherein the second vocabulary comprises a set of tree structured clusters of binary descriptors; comparing, at the advertisement detection system, each spatially identifiable binary descriptor from the pre-defined number of binary descriptors corresponding to the first vocabulary of each selected frame with a plurality of spatially identifiable binary descriptors in at least one or more clusters of the set of tree structured clusters corresponding to the second vocabulary of the binary descriptors of the repository of the ad frames; progressively scoring, at the advertisement detection system, each selected frame from each live stream of the media content for validation of the selected frame as the ad frame of a first ad, wherein the first ad being progressively scored for each positively validated frame to obtain a progressive ad score and wherein the progressive score being calculated for the ad length independent detection of the first ad; and detecting, at the advertisement detection system, the first ad in each live stream of the media content of the one or more broadcasted channels when the progressive score for the first ad being above a threshold score, wherein the first ad being broadcasted by the one or more broadcasted channels in one or more languages and wherein the first ad being detected independent of the one or more languages of the first ad broadcasting in the one or more broadcasted channels.

2. The computer system as recited in claim 1, further comprising assigning, at the advertisement detection system, a weight value to each binary descriptor of the pre-defined number of binary descriptors, wherein the weight value corresponding to each binary descriptor being L1 normalized and wherein each normalized weight value corresponding to each binary descriptor being characterized by an arithmetic sum of 1.

3. The computer system as recited in claim 1, wherein the first ad in each live stream of the media content being detected in a detection period, wherein the detection period being in a range of 0.6 second to 1 second and wherein the detection period being equivalent to receiving of two frames of the first ad.

4. The computer system as recited in claim 1, wherein the binary descriptors in the first vocabulary being compared with the binary descriptors in the second vocabulary by: searching for a cluster in the set of tree structured clusters comprising of the binary descriptors with minimum hamming distances with the corresponding binary descriptors of the selected frame in corresponding live stream of the media content; matching each binary descriptor from the first vocabulary of the binary descriptors for the selected frame of the media content with the binary descriptor in the searched cluster for the minimum hamming distance; calculating Bhattacharya distance between weight values of each matched pair of the binary descriptors; and summing calculated Bhattacharya distance for each descriptor in the first vocabulary of the binary descriptors corresponding to the selected frame of the media content to obtain a summed feature value, wherein the summed feature values being in a normalized range of 0 to 1.

5. The computer system as recited in claim 1, further comprising receiving, at the advertisement detection system, the live stream of the media content of the broadcasted channel, the live stream of the media content comprises of a live ad stream and a non-ad stream and the live ad stream comprises one or more ads.

6. The computer system as recited in claim 1, wherein each selected frame in the selected set of frames differ by a frame gap of at least 8 frames, wherein the set of frames selected from the pre-defined set of frames in the live stream of media content per second being 3 and the pre-defined set of frames per second in the live stream being 25 and wherein the pre-defined number of keypoints being 700, the pre-defined number of descriptors being 700 when the pre-defined number of keypoints being 700 and the pre-defined number of spatially identifiable binary descriptors being 2100.

7. The computer system as recited in claim 1, wherein the second vocabulary of the binary descriptors being characterized by an n-ary tree data structure comprising of leaf nodes, the second vocabulary of the binary descriptors being created by: extracting the pre-defined number of keypoints and corresponding binary descriptors from each frame of the repository of the ad frames; creating the special pyramid of the descriptors for each ad frame to obtain the pre-defined number of spatially identifiable binary descriptors; clustering the binary descriptors into a first set of clusters, the binary descriptors being clustered into the first set of clusters based on an evaluation of a minimum hamming distance between each binary descriptor; iteratively clustering the binary descriptors in each cluster of the first set of clusters and each subsequent cluster for a pre-determined iteration level to obtain a set of tree structured clusters; and assigning a weight value to each clustered binary descriptor based on a term frequency and an inverse document frequency and normalizing the weight values using L1 normalization, the weight value being normalized for an evaluated arithmetic sum of weight values as 1.

8. The computer system as recited in claim 7, wherein each cluster in the first set of clusters and the set of tree structured clusters being associated with a centroid, wherein the first set of cluster comprises 10 clusters and corresponding 10 centroids and the second set of tree structured clusters comprises 106 clusters and 106 centroids and wherein the pre-determined iteration level for clustering binary descriptors corresponding to each frame of the repository of ad frames being 6.

9. The computer system as recited in claim 1, wherein each selected frame for the first ad in the corresponding live stream of the media content being progressively scored in at least one or more steps, t he one or more steps comprises: comparing the summed feature value for each selected frame with a first threshold value for validating the selected frame as the ad frame of the first ad; evaluating a ratio test for determining degree of difference between each selected frame in the selected set of frames in the live stream of the media content, wherein the ratio test being evaluated by dividing the summed feature value for a second frame by the corresponding summed feature value for a first frame in the selected set of frames; rewarding the first ad of one or more ads in the live stream of the media content by assigning a first ad score for a positive validation of the evaluated ratio, wherein the first ad score being assigned to the first ad when the ratio being less than a second threshold value; penalizing a second ad of the one or more ads in the corresponding live stream of the media content by deducting a second score from the assigned first ad score for the second ad, wherein the second ad being a past ad streamed before the first ad and wherein the first ad being streamed in real time in the corresponding live stream of the media content; rewarding the first ad of the one or more ads in the corresponding live stream of the media content by adding a third score to the first ad score of the first ad, wherein the third score being rewarded based on an equality criterion and wherein the equality criterion being based on equality of the feature value of the first frame and the feature value of the second frame in the selected set of frames; rewarding the first ad of the one or more ads in the corresponding live stream of the media content by adding a fourth score to the first ad score of the first ad, wherein the fourth score being rewarded based on a vicinity criterion and wherein the vicinity criterion being based on successive positive validation of the selected set of frames; and calculating the progressive ad score for the first ad and the second ad based on at least one of progressive addition and subtraction of the second score, the third score and the fourth score from the first ad score.

10. The computer system as recited in claim 9, wherein the first threshold value for validating comparison of the summed feature value being 0.02 and the second threshold value for validation of the evaluated ratio being 0.7 and wherein the first ad score for positive validation by the ratio test being 1.5, the second score for penalizing the second ad being 0.5, the third score for the equality criterion being 0.5 and the fourth score for the vicinity criterion being 0.5.