JPH03184484A - Commercial detector - Google Patents

Abstract

PURPOSE:To decrease erroneous detection and to improve accuracy for detection by calculating AND between the detection output of a silent part an a CM detection output by a comparing means since there is this silent part for 0.5-2.0 seconds almost without fail at the beginning and end of a CM. CONSTITUTION:A silent part detector 26 detects the silent part for 0.5-2.0 seconds provided almost without fail at the beginning and end of the CM. The output of the silent part detector 26 is inputted to an input end 29 of a gate 27. At the gate 27, the output of a comparator 24 is outputted to an output terminal 30 for fixed time after the output of the silent part detector 26. When there is no output form the silent part detector 26, the gate 27 executes no output even incase there is an output form the comparator 24. This gate output is the CM detection output. Since this AND output is defined as the CM detection output, the output of the comparator 24 can be deleted at the time when a broadcasting state is a main program. Namely, the erroneous detection of the output form the comparator 24 is decreased and the detection accuracy can be improved.

Description

DETAILED DESCRIPTION OF THE INVENTION [Object of the Invention] (Industrial Field of Application) The present invention provides a commercial detection output useful for use in household image recording and reproducing devices such as VTRs.

(Conventional technology) Generally, in television broadcasting, commercials such as ↑
・Le (hereinafter referred to as CM) is acting as a matchmaker. On the other hand, for viewers, commercials do not necessarily provide necessary information;
When recording on a TR or the like and using television broadcasting, it is advantageous to be able to record and play back without commercials.

However, until now, there has been no device that provides such a deletion function. The main reason why such a function is not provided is that it is difficult to judge whether a received signal is a CM or not.

Conventionally, the method for detecting commercials is to use commercials obtained in advance.
There is a method of comparing the characteristics of M and the characteristics of the CM being broadcast to see if they match.

FIG. 5 shows an example of a conventional CM detector, and FIG. 6 shows a specific location on the screen from which broadcast image data is to be extracted.

In FIG. 5, the broadcast side is inputted to the broadcast image data extractor 2 from the input terminal 1. The broadcast image data extractor 2 divides the screen into several blocks of nxm pixels, such as specific locations 3 and 4, as shown in Figure 6, takes the average brightness of the nxm pixels in each block, and obtains data that identifies the broadcasting side. Created.

On the other hand, in the 0M data memory 5, data for specifying a large number of CM pictures is created and input in advance using the same method as the broadcast picture data extractor 2.

Then, the output of the broadcast image data extractor 2 is inputted to one input terminal 6 of the comparator 8, and the output of the 0M data memory 5 is inputted to the other input terminal 7 of the comparator 8.

The comparator 8 compares the output of the broadcast image data extractor 2 and the output of the 0M data memory 5, detects whether the two outputs match or not, and outputs it from the output terminal 9.

FIG. 7 shows the configuration of the comparator 8.

In the comparator 8, the absolute value I A-B l of the difference between the input A at the input end 7 and the input B at the input end B is obtained by the absolute value difference detector 10, and the absolute value I A-B + is obtained by the adder. 11 and memory 1
2 for one screen, and the comparator 13 compares whether the total value is less than a certain value, and outputs a match output when it is less than a certain value.

- Once a rescue request is obtained, the broadcaster (i.e. CM) with the same content
It is assumed that a commercial has been detected.

In general, with the methods described above, increasing the amount of data extracted from the screen reduces false positives and improves accuracy, but the memory capacity
．． There is a problem in that the circuit scale of the comparator 13 and the like increases, which increases cost. Conversely, reducing the amount of data extracted from the screen will reduce the circuit size and cost, but will also increase false detections and reduce detection accuracy. Therefore,
The above method has not been put into practical use due to the relationship between detection accuracy and cost.

Furthermore, since the above method cannot detect CMs that have not been input to the 0M data memory 5, it is essential to record the screen data of the CMs in the 0M data memory 5 by some method and construct a library. Moreover, since new commercials are added every day, it is also necessary to update the library in the 0M data memory 5.

However, conventionally, there has been no method for rationally inputting and updating this 0M data.

(Problem that the invention is trying to solve) As mentioned above, in the conventional method, detection accuracy is determined by each data extracted from the screen, but increasing the detection accuracy increases the circuit scale, and it is difficult to reduce the circuit scale. Otherwise, the detected viscosity will drop.Furthermore, there is no rational way to construct a library of data that specifies the CM screen in the 0M data memory.

Therefore, the present invention aims to eliminate the above problems.
It is possible to increase the detection accuracy without reducing the circuit size, and at the same time, it is possible to record data that specifies the CM screen in the 0M data memory and build a library, and update the 0M data memory when there is a new CM. The object of the present invention is to provide a commercial syringe detection device that can perform the following steps.

[Structure of the Invention] (Means for Solving One Problem) The commercial detection device of the present invention includes a feature extraction means for extracting features of a received image, and a feature extraction means for extracting features of a plurality of commercial images in the same manner as the feature extraction. A storage means for extracting and storing the information,
Comparison means for comparing the outputs extracted by the feature extraction means and the storage means and detecting a match between the two outputs; The present invention is characterized by comprising: a detection means; and a gate means for performing a logical product of the output of the comparison means and the output of the non-speech detection means to obtain a 0M detection output.

Furthermore, in addition to the above configuration, there is provided means for manipulating the address of the 0M image data that matches the received image data when a CM is detected, and sorting the CMiij image data in the order of the most recently received; and the silent part detecting means. Even if there is a detection output, if there is no 0M image data matching the received image data in the storage means, by providing means for recording the current received image data in the storage means as new 0M image data, the library can be stored. It is possible to construct, record and update new 0M image data.

(Function) In the present invention, 0.5 to 2
Since there is almost certainly a non-speech portion of about seconds, by ANDing this non-speech portion detection output with the 0M detection output by the comparison means, false detections can be reduced and detection accuracy can be increased.

In addition, since the start time of a new CM, which cannot be detected using conventional methods, can be detected by silent part detection, new 0M
At the same time as the detection, a new CM can be imported and the storage means can be updated.

(Example) Examples will be described with reference to the drawings.

FIG. 1 is a block diagram showing a commercial detection apparatus according to an embodiment of the present invention.

In this figure, commercial detection device 1? ? It, 0M data memory 21, broadcast image data extractor 23, comparison X2/1, silent part detector 26, gate 27
It is composed of.

The comparator 24 corresponds to the comparator 8 of the conventional example shown in FIG.
The comparison output is inputted to the input terminal 28 of the gate 27. The broadcast picture data extractor 3 corresponds to the conventional broadcast picture data extractor 8 shown in FIG. Create data that identifies broadcast images by taking the brightness average. On the other hand, the broadcast audio is input to the silent part detector 26. Silent part detector 26
detects 0.5 to 2 seconds of silence, which is almost certainly at the beginning and end of a commercial. Through this detection, it is possible to know the start and end times of commercials. Silent part detector 2
6 is input to the input terminal 29 of the gate 1-27. The gate 27 outputs the output of the comparator 24 to the output terminal 30 for a certain period of time after the output of the silent portion detector 26 is received.

When there is no output from the silent part detector 26 (the comparator 24
Even if there is an output, the gate 27 does not output.

This gate output is the 0M detection output in this embodiment.

FIG. 2 is a timing chart for explaining the operation of the gate 27. In FIG. 2, (a) shows a broadcast state, in which a plurality of commercials are mixed in the main story. (b
) shows the output of the comparator 24. This output (a3) is obtained by correlation comparison of the screen data, and includes an output that matches the 0M data even during the main part.

(C) is the output of the silent part detector 26, which detects silent parts at the beginning and end of a commercial. Then, as the output of the gate 27, b is obtained by ANDing the output of (b) and the output of (C). By using this AND output as the 0M detection output, it is possible to eliminate the output of the comparator 24 at the ++ and y intervals where the broadcasting state is the tree line. This makes it possible to reduce erroneous detections of the comparator 24 output and improve detection accuracy. d3, in the explanation of Figure 2, 0
The M data memory 21 stores the first screen of a CM as 0M data. Therefore, the comparator 24 output is provided at the beginning of the CM, and the gate 27 output is also obtained at the beginning of the CM. Therefore, from the output of the gate 27, only the starting point of the CM can be known, but the ending point cannot be known. However, most of the commercials are 15 seconds long.
It lasts 30 seconds. Therefore, if the fame output is detected 15 seconds after the gate 27 output and there is no voice part, the output will be 15 seconds.
It can be determined that this is the end point of a second-long commercial, and if there is an audio output 15 seconds after Game 1 receives the 27-second output, it can be determined that it is a 30-second commercial.

On the other hand, in FIG. 1, the 0M data memory 21 is composed of a memory manipulator 31, an address generator 36, and a memory 39. At the same time, the output of the silent part detector 26 is input to the input terminal 33 of the memory operation device 31. If there is human power, the memory operation signal A is output from the output terminal 34.
The memory operation device 31 outputs the memory operation signal B from the output terminal 35 when the gate output is not input from the input terminal 32 within a certain period of time after the input is input to the input terminal 33.

The memory operation signals A and B are then input to input terminals 37 and 38 of the address generator 36.

3 and 4 show the memory 3 of the address generator 36.
9 is an explanatory diagram illustrating an address operation for 9. FIG.

FIG. 3 shows an address operation for arranging CM data in the order of newest.
2,3. -To-CM data A, B, C.

If . Assuming that the CM is detected again, a memory address operation is performed to place the CM data C at the lowest address as shown in Figure 3(b).Then, the CM data C is placed at the lowest address. A memory address operation is performed by incrementing the addresses of CM data A and CM data B one by one, which are younger than the address of the previous CM data C, and setting them as address 2 and address 3. Assign the youngest addresses in the order in which they are broadcast.

Figure 4 shows the address operation for updating new 0M data.As shown in Figure 4(a), addresses 1, 2, 3,
．． ...... CM data A, B, C, ......
is compatible, the memory operation signal B is output from the input terminal 38.
is input, it is assumed that a new CM has been detected, and the broadcast image data input to the memory 39 from input #a40 is recorded as CM data, as shown in FIG. Perform address operations for memory allocated to the younger address 1. Then, the addresses of all CM data other than the broadcast image data just recorded as CM data in the menu 39 are incremented one by one. When the amount of CM data exceeds the capacity of the memory 39, the CM data (CM data Z in the figure) that has not been broadcast recently at the oldest address disappears. Through this series of address operations, a new CM
Record and update data.

The amount of data of the CM data mentioned above is 6X4=2 when the average brightness of each block at 6 locations on the screen is expressed in 4 bits.
It becomes 4 bits. There are 224 types of CM data that can be expressed with this amount of data. Also, since there are about 2000 types of commercials that are being broadcast, the number of types of commercial data that can be recorded in the 0M data memory 21 is practically 1000.
! ! It's good if there is something similar.

Now, let us consider the case of performing CM detection using the conventional method shown in FIG. Assuming that the tIi image sending data appears randomly, =4.7 [minute 1] That is, there will be an erroneous detection once every 4.7 minutes.

Now, considering that similar CM detection is performed in the embodiment of the present invention, when the silent part detection correctly detects a CM, no false detection occurs, so only when the silent part detection causes a false detection. Just think about it. The false detection of silent part detection is 24
As a result of receiving and observing broadcasts for 0 minutes, it was about 20 times per 30 minutes.

After detecting a silent part, the output of the comparator 24 is sent to the gate 27.
If the constant time for output is 1 second, then =7.0 [hours] That is, erroneous detection will occur only once every 7.0 hours.

Compared to the conventional example, the number of false detections can be reduced to 1790.

[Effects of the Invention] As described above, according to the present invention, false detections can be significantly reduced and detected viscosity can be increased.

Furthermore, it is possible to automatically build a library by recording data specifying a CM screen in the 0M data memory, and it is possible to automatically update the 0M data memory when a new CM is received. Furthermore, the number of circuits to be added can be reduced.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing a commercial detection device according to an embodiment of the present invention, and FIG. 2 is a timing chart 1-1 explaining the operation of the gate in FIG. 1. FIG. 5 is a block diagram illustrating a conventional commercial detection device; FIG. 6 is an explanatory diagram illustrating the operation of a broadcast image data extractor; FIG. 7 is a block diagram illustrating a comparator. It is. 21...0M data memory, 23...Broadcast image data extractor, 24...Comparator, 2
6... Silent part detector, 27... Gate, 31...
・Memory operation device, 36...Address generator, 39...
·memory. Figure 2 (0) +1)) Figure 5 Procedure? = Hoshosho (spontaneous) Furushu January 24, 1990

Claims (1)

[Scope of Claims] A commercial detection device for detecting whether a received image is a commercial or a main program in a device that receives a television broadcast signal, comprising: a feature extraction means for extracting features of the received image; Similarly, a storage means for extracting and storing features of a plurality of commercial images, and a comparison means for comparing outputs extracted by the feature extraction means and the storage means and detecting a match between the two outputs. , a silence detection means for detecting a voiceless part of the received voice and knowing the start point and end point of a commercial, and a logical product of the output of the comparison means and the output of the silence detection means to obtain a commercial detection output. A commercial detection device comprising: a gate means for obtaining a commercial.