JPH11112922A - Method and device for stream event point detection display - Google Patents

Abstract

PROBLEM TO BE SOLVED: To display the present detection condition to a user in an understandably manner by inputting a leading section of an input stream to a read-ahead buffer, detecting whether or not there is an event point in the look-ahead buffer, fast forwarding an event point detection processing of each stream according to the presence or absence of the event point or doing the same by switching to a normal speed. SOLUTION: A frame sequence 221 from a current read-in position to a next I picture is read in a read-ahead buffer 203 from a video stream input device 201 and a fast-forward event point detection device 207 discriminates whether or not a scene change point exists in the read-ahead buffer 203. When a scene change point exists, the frame position of a scene change event point is detected by using a detail event point detection device 205, and when it is detected, the detection result is displayed on a display area. When the scene change point does not exist, the frame display an image L2 on the same area and repeats this until there is no video.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an apparatus for analyzing the contents of stream data such as video, audio, and musical score data and the apparatus, and more particularly, to a scene change point at which a camera image is switched and a speaker of radio sound. The present invention relates to a method and an apparatus capable of efficiently detecting an event point on a stream such as a switching speaker change point.

[0002]

2. Description of the Related Art With the recent development of computer technology,
A large amount of data in a stream format such as video data, audio data, musical score data called MIDI, and additional information data strings such as subtitles attached to a video can be handled in a large amount and flexibly. In the future, a technology that allows a user to easily manage such a large amount of stream data will be important. As one of the stream data management technologies, a scene change point indicating a place where a TV image is switched to a CM in a stream, a speaker change point indicating a point at which a speaker is switched by radio sound, and the like are included. There are techniques for detecting an event point representing a certain event in a stream. In addition, the event point may be a time when the rhythm of the music changes with respect to the music score data stream,
For additional information data strings such as subtitles attached to the video,
For example, when a certain keyword appears.

As a method of detecting the scene change point described above from video data, there is Japanese Patent Application Laid-Open No. 8-227462. A method for detecting a point in time when a dissolve special effect, which is one of video special effects, appears.
-65287. In each of the methods described in the related arts, a frame image included in a video is read out one by one, and a scene change point or a dissolve point is determined based on an image difference amount between the frame image and a frame image near the frame image. This is a method of detecting a special effect point.

[0004] As a method for finding a change point of audio data such as a transition from a state close to silence to a sound state such as the beginning of music, a method used in an audio dubbing apparatus used for an audio tape, a mini disk, or the like is used. The following methods are known in the art. In this method, the average of the amplitude values of the sound intensities is averaged at regular time intervals, and if the average within the time exceeds a certain threshold value, it is determined that the sound is present. Has found a turning point.

However, as a problem common to the above methods, it is necessary to refer to all the frame images of the video data stream or to refer to all the audio waveforms of the audio data stream. Event points cannot be detected in a shorter time than the original playback time.

In Japanese Patent Application Laid-Open No. Hei 9-107517, in order to solve the above-mentioned problem, the detection processing time is shortened by using the following method of detecting a scene change point while reproducing a video at high speed. ing. In this change point detection control method, first, a frame image is fetched at a fixed time interval while reproducing a video at high speed as a first detection pass, and the difference between the newly fetched frame image and the previously fetched frame image is greatly different. Listed as event point candidate sections (scene change point candidate sections). Next, as a detection second pass, each of the detected video event point candidate sections is detected one frame at a time by reproducing at a normal speed, and an accurate event point position is detected.

[0007]

The above-described conventional method has the following problems.

The first problem is that the conventional method disclosed in Japanese Patent Application Laid-Open
In the technique described in Japanese Patent No. 7517, it is difficult for a user to understand which part of the input video is being subjected to the detection processing. The reason will be described below.

In the change point detection control method, first, a detection first
In the pass, a screen that scans the entire input video is displayed like a fast-forward video. Next, in the second detection pass, the video of the video event point candidate section detected at the time of the fast-forward playback is played back continuously for each candidate section. Because of such a two-stage detection method, the video is reproduced as a fragmented video different from the original reproduction order, and it is difficult to know which part of the video is currently being detected.

[0010] The second problem is that in the change point detection control method, when the user terminates the video detection process, the first detection is performed.
During the pass, only the video event point candidate section remains, so that the position of the event point with one elementary data accuracy cannot be detected.

The third problem is that the change point detection control method uses the video / audio data or T
In the case of stream data, such as video and audio data received as V broadcast, having the characteristic that the stream is deleted once it is output to a TV or radio, etc., the second pass of detection cannot be executed, so that one elementary data Event points with unit location accuracy cannot be detected.

The fourth problem is that in the case of the method of detecting a special effect point of a video or the method of detecting a sound change point described in the conventional method, the change point detection control method cannot be applied.

A fifth problem is that in the case of the change point detection control method, when there is an erroneous detection, for example, a point to be detected cannot be detected or a point that need not be detected is detected, a detection process is performed. It is difficult for the user to control the contents of the detection. This is directly attributable to the fact that the current detection situation shown in the first problem is difficult for the user to understand.

A first object of the present invention is to provide a fast-forward type stream event point detection method for detecting an event point from stream data such as video and audio with a small amount of calculation, so that the current detection status can be easily understood by a user. It is to display.

A second object of the present invention is to provide a method for detecting a stream event point in a fast-forwarding system, in which an event point having a position accuracy of one raw data unit can be left even if the user terminates the detection processing on the way. That is what.

A third object of the present invention is to provide a fast-forwarding system which has the characteristic that when a stream is once output to a TV or the like, such as the above-mentioned TV broadcast stream data, the stream data is erased. An efficient detection method such as a stream event point detection method is to detect an event point having a position accuracy of one elementary data unit.

A fourth object of the present invention is to provide a method for efficiently detecting a video dissolve special effect point or an audio change point, such as the fast-forward stream event point detection method. Is to make the method applicable.

A fifth object of the present invention is to allow a user to easily control the contents of detection in the fast-forward type stream event point detection method during the detection processing.

[0019]

The present invention uses the following method to achieve the above object. FIG. 1 is a flowchart showing an outline of the method of the present invention. In the following description, (step xxx) indicates a step number in the schematic flow of FIG. FIG. 2 is a system configuration diagram of a stream event point detection / display device using the method of the present invention. In the following, the method of the present invention will be described first with reference to FIG. 1, and finally the outline of the stream event point detection display device using FIG. 2 will be described.

In order to achieve the first to fourth objects of the present invention, event points such as a scene change point of a video and a speaker change point of a voice are detected from input stream data such as a video and a voice. The method for detecting and displaying stream event points to be output is characterized in that the following processes (1) to (5) are performed.

(1) The head section is input to the prefetch buffer (step 1102).

(2) It is detected from the prefetch buffer whether or not an event point exists in the prefetch buffer (step 1103).

(2a) According to the event point detection presence / absence information obtained in (2), either (3) or (4) is performed (step 1104).

(3) If it is determined in (2) that an event point exists, all elementary data (minimum units constituting a stream) in the stream section in the prefetch buffer are sequentially read, and the following (3a) and (3b) The processing of (3c) is performed (step 1111).

(3a) The raw data is output to the user (step 1121).

(3b) It is determined whether or not the position of the raw data is an event point by using the feature amount of the raw data (step 1122).

(3c) Only when it is determined in (3b) that the position of the elementary data is an event point, (3d)
(Step 1123).

(3d) Using the position of the raw data as an event point position, a detection process such as recording of the event point position is performed (step 1131).

(4) If it is determined in (2) that there is no event point, the stream in the prefetch buffer is fast-forward displayed (step 1112).

(5) These steps (1) to (4) are sequentially repeated (step 1101).

By the above method, the first object of the present invention can be solved for the following reasons. Where the event point does not exist in the stream as in (4), the stream can be fast forward reproduced as in the prior art. In addition, as shown in (3), in the case where an event point is likely to exist in the stream, detailed stream data is converted to one elementary data without indicating to the user a stream rewind output that disrupts the reproduction order of the stream. Can be shown slowly to the user. As described above, unlike the related art, the method of the present invention can obtain an output that is easy for the user to understand by automatically switching the output speed of the stream according to the content of the stream.

According to the method of the present invention, for the following reasons:
The second object of the present invention can be solved. In the present method, unlike the conventional method, event point candidate points are not detected, and in detecting each event point, an event point position having an accuracy of one elementary data unit is detected. For this reason, in the above method, even if the processing is terminated at any point during the detection processing, the event point position with one elementary data accuracy can be calculated.

According to the method of the present invention, for the following reasons:
The third object of the present invention can be solved. In the above method,
As in (2), the stream data is always read into the prefetch buffer, and as in (3), the event buffer is used when detecting an event point in units of one elementary data. Therefore, even in the case of stream data, such as stream data of a TV broadcast, which is deleted when the stream is once output to a TV or the like, in the method of the present invention, since the video is buffered, the data is stored in units of one raw data. Event points with positional accuracy can be detected.

According to the method of the present invention, for the following reasons:
The fourth object of the present invention can be solved. For example, in the case of the detection of the above-described dissolve special effect point of the video, in (2), when the image difference amount between the frame image in the current look-ahead buffer and the previous frame image in the look-ahead buffer is large, it is determined that an event point exists. Judge, and in (3),
The dissolve detection method disclosed in Japanese Patent Application No. 7-210409 may be used. This is because the dissolve looks like the scene change point when the image is a fast-forward image. Also, as for the voice, for example, in (2), voice waveform data in which the voice waveform in the prefetch buffer is thinned out every 2 seconds to a length of 1 second, the amplitude average of the voice waveform is obtained, and the value is obtained. If the threshold value is exceeded, it is determined that there is an event point, and in (3), the conventional method may be used as it is.

Here, an apparatus for achieving the fifth object of the present invention will be described.

Based on the above processes (1) to (5), if the following corrections are made, detection control that is easy for the user to understand can be realized.

For example, when the user wants to forcibly detect an event point at the stream position currently being detected by a button or the like meaning a forced detection during the detection process, the above (3c) and (4) are described below. Change to Of course, the present example can be implemented by an implementation method other than the following.

In (3c) and (3b), when it is determined that the position of the elementary data is an event point, or when the user presses the forcible detection button, detection of the recording of the event point position, etc. Time processing.

(4) If it is determined in (2) that there is no event point, and if the user has pressed the forcible detection button, detection time processing such as recording of the event point position is performed. In this case, the stream section in the prefetch buffer is displayed by fast-forwarding the stream.

As another control method, for example, when the user ignores the process of detecting the position of an event point with one elementary data accuracy by using a button or the like meaning that detailed detection is ignored during the detection process, the above-mentioned (3) Is changed as follows. Of course, the present example can be implemented by an implementation method other than the following.

(3) If it is determined in (2) that an event point exists, the head element data position at the head of the prefetch buffer is set as the event point position, and a detection process such as recording of the event point is performed. Fast forward display of the stream section in the buffer.

Finally, the outline of the stream event point detecting and displaying apparatus of the present invention will be described with reference to FIG. FIG. 2 is a system configuration diagram of a stream event point detection display device using the method of the present invention.

The stream event point detection control of the present invention
The display device includes the following devices. This apparatus can execute processing according to the above-described stream event point detection and display method.

The stream event point detection control of the present invention
The display device includes a stream input device 1001 for reading an input stream, a look-ahead buffer 1102 for storing a head section of the read input stream, a stream output device 1008 for outputting a stream to a user or the like, and a look-ahead buffer. Fast-forward output control device 1003 for outputting a stream in fast-forward format 1002
A detailed output control device 1004 for outputting the stream in the prefetch buffer 1002 one by one data, and a fast-forward event point detection device 10 for judging the presence or absence of an event point in the stream in the prefetch buffer 1002
05, the detailed event point detection device 1006 that determines the position of the event point in the stream in the prefetch buffer 1002 with one elementary data precision, and the fast-forward event point detection device 1005 has no event point in the prefetch buffer 1002. When it is determined that the content of the prefetch buffer 1002 is fast-forward output using the fast-forward output control device 1003, and when the fast-forward event point detection device 1005 determines that there is an event point in the prefetch buffer 1002,
The raw data in the prefetch buffer 1002 is sequentially read out, and the detailed output control device 1008 displays the raw data, and the detailed event point detection device 1006 determines whether there is an event point at the position of the raw data. Which is characterized by having a detection control device 1007 that performs

[0045]

DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to FIGS.

FIG. 3 is a system configuration diagram for detecting a scene change point, which is a type of event point, from an MPEG file. In the following, an “MPEG file” is an MPEG file defined by the international standard ISO117172-1.
1, or a video data file stored in MPEG2 or a video video format equivalent thereto.

An outline of a processing flow in this embodiment will be described with reference to a system configuration diagram of FIG. 3 and an execution screen of FIG.

In this embodiment, (1) the current reading position (F2 in FIG.
A frame sequence 221 from 1) to the next I picture (a frame image meaning key element data in an MPEG file video stream. At the time of FIG. 3, the position F23 is set to an I picture), a frame sequence 221 is read into the prefetch buffer 203, and (2) The fast-forward event point detection device 207 determines the presence or absence of a scene change point in the look-ahead buffer. If a scene change point exists in (3) and (2), all frames in the look-ahead buffer 203 are displayed. Detailed event point detection device 20
5, the frame position of the scene change event point is detected, and if detected, the detection result is displayed in the result display area 302 of FIG.
(In the system configuration of FIG. 3, the video stream output device 2
13), and if there is no scene change point in (4) and (2), the frame image L2 is displayed in the same area, and (5) (1) to (4) are displayed as images. Repeat until there is no more.

FIG. 4 shows an example of an execution screen of the high-speed video event point detection method. An area 301 is an input video display area, and displays which frame image of the input video is currently actually being detected. Button 304, 30
Reference numerals 5 and 306 denote detection control buttons.
When 04 is clicked with a user operation input device such as a mouse, the detection process shown in the process flow of FIG. 5 is started.
When the detection end button 305 is clicked during the execution of the detection processing, the detection processing is terminated. When the forcible detection button 306 is clicked, the currently displayed frame image and frame position are forcibly recorded as video event points.

In the area 302, a list of detected video event points is visually displayed. Icon 310,
Numerals 311 and 312 display the frame positions of the detected video event points and the frame images at the frame positions. For example, the icon 312 represents a video event point at the frame position 100, and a frame image at the frame position and a numeral 100 representing the frame position are displayed.

A check box 307 indicates designation of a detail detection mode, and is in a checked state at the time of initial display. Clicking the check box toggles the checked state and the unchecked state. When the check box is checked, the detection position accuracy of the event point is one frame unit, and when the check box is not checked, the detection position accuracy is a coarse detection accuracy of two or more several frame units.

The slider 308 displays the current reading position of the input video data when no detection is performed.
By moving the knob 309 of 08, the reading position of the input data can be moved. In the input video display area 301, a frame image of the position specified by the slider 308 in the input video data is displayed. In this example, the slider 308 cannot be used during the detection process.

When the play button 321 is clicked, the video is played. Clicking the stop button 322 stops video playback.

FIG. 5 shows a processing flow when the detection start button 304 in FIG. 4 is pressed.

In step 401, the video prefetch buffer is initialized. In practice, the video prefetch buffer is emptied, and a black (all pixel values are 0) frame image is substituted for a variable LastFrameFF for storing the frame image.

In step 402, initialization of fast-forward event point detection is performed. For example, a process such as painting a previous frame image as an internal variable into black is performed.

Step 403 and steps 411 to 4
At 17, a loop process for repeating a process of detecting a video event point from a video is performed. In one loop, video event point detection processing is performed on the frame image sequence read into the prefetch buffer. The condition for ending the loop is when all the images have been read, or when the user presses the end button in the loop.

In step 411, the first several frames of the input video data are read out and written into the prefetch buffer. In the present embodiment, a frame image sequence from the current read frame position to the position of the next appearing key frame is read into the prefetch buffer. For example, a key frame in the case of an MPEG file means an I picture that exists in video data every 10 to 30 frames.

When a prefetch buffer is implemented, all the still image data of the frame image sequence for which the read instruction has been issued may not be recorded in the prefetch buffer, and the input image data of the frame image sequence for which the read instruction has been issued may be omitted. Only the first frame position and the last frame position are recorded, and the relevant frame image may be read from the input video data only when an instruction to refer to the frame image from the video prefetch buffer is issued. By using such a delayed reading method of the prefetch buffer, it is possible to save both the memory capacity of the prefetch buffer and the frame image acquisition time from the prefetch buffer. For example, if the input video data is an MPEG file, the current read frame position and the next I
The position of the picture and the frame image obtained by expanding the I picture are recorded in the prefetch buffer, and when the frame image data is actually extracted from the prefetch buffer, if the extracted position is the position of the I picture, the frame image data in the prefetch buffer is read. Is directly extracted, and if the extraction position is other than the position of the I picture, the corresponding frame image data is extracted from the input video data.

In step 412, the key frame image recorded at the last position of the prefetch buffer is extracted and stored in aFrameFF.

In step 413, fast forward event point detection processing is performed. In this embodiment, in order to detect the presence or absence of a scene change point, the frame image aFrameFF and the frame image Last passed to the previous fast-forward event point detection function are used.
The image difference amount from FrameFF is calculated, and if the image difference amount is larger than a certain threshold value, it is determined that there is a scene change point.

In the conditional branch of step 414, if it is determined in step 413 that there is an event point and if the current detection state is the detailed detection mode, steps 421 and 42
2, an event point detection process is performed for each frame. Here, the detailed detection mode indicates a state in which the detailed detection check box 307 in FIG. 2 is checked.

In the conditional branch of step 414, step 4
13, it is determined that there is an event point, but if the current detection state is not the detailed detection mode, step 423
, The detection point processing is performed by setting the position of the event point as the position of aFrameFF and the image of the event point as the frame image of aFrameFF. The details of the detection process will be described in detail in step 443.

If the conditional branch in step 414 does not satisfy any of the above two conditions, in step 415,
The frame image aFrameFF at the time of fast forward is displayed.

At 421, an initialization process for detecting a detailed event point is performed. Here, the variable La that stores the frame image
A black (all pixel values are 0) frame image is substituted for stFrame.

In addition to the present embodiment, for example, when the method of detecting a change point of a moving image disclosed in Japanese Patent Application Laid-Open No. 8-227462 is directly applied to the detailed event point detection process, the detection method is used. Cannot detect a scene change point for the first two frames of a given frame sequence.
What is necessary is just to input LastFrameFF twice to the method.

In steps 422 and 431 to 435, the following detailed event point detection processing is performed on the frame image sequence in the prefetch buffer.

In step 431, the frame image at the i-th frame position in the prefetch buffer is acquired as aFrame.

In step 432, the frame image aFrame
Is displayed in the input video display area 301 of FIG.

At step 433, processing for one frame for detecting a detailed event point is performed. In this embodiment, in order to detect the presence or absence of a scene change point, a frame image aFrame
Then, the image difference amount from the previous frame image LastFrame is calculated, and if the image difference amount is larger than a certain threshold value, it is determined that there is a scene change point. In addition to this implementation example, for example, the detailed event point detection processing is performed by using the conventional technology disclosed in
The change point detection method of a moving image disclosed in Japanese Patent Application No. 227462 may be directly applied.

In the conditional branch of step 434, if it is determined in step 433 that an event point exists, it is determined that the event point is detected at the position of the frame image aFrame. In step 443, the event point is stored at the head of the prefetch buffer. The number of the frame position in the input video data corresponding to the present frame image and the loop counter i is added to the detection position of the event point, and aFrame is set to the event point frame image, and the detection process is performed.

In the detection process, an event point frame image and an event point detection position are additionally displayed in the form of an icon like an icon 312 at the end position of the icon list displayed in the area 302 of FIG.

In step 435, a condition is determined based on the user operation event obtained by the user operation event input device. If the user operation event is the pressing of the forced detection button 306 in FIG. 2, the event point position is determined in step 441. The position of the frame image aFrame is set, and the event point frame image is set to aFrame. The same detection processing as in step 443 is performed. If the detection end button 305 is pressed, the process exits the inner loop again in step 442 to perform the processing. Move to step 416, otherwise do nothing.

In step 416, a condition is determined based on the user operation event obtained by the user operation event input device. If the user operation event is the pressing of the forcible detection button 306 in FIG. Is set to the frame position of aFrameFF, the event point frame image is set to aFrameFF, and the same detection processing as in step 443 is performed. If the detection end button 305 is pressed, the current loop is exited in step 424 and the entire detection processing is ended. And do nothing else.

In step 417, the current fast-forward frame image aFrameFF is substituted for the previous fast-forward frame image LastFrameFF.

FIG. 6 shows an example of a method for visually displaying the detection status using the detection position slider.

In this example, whether the detection processing is currently detecting fast-forward or detailed detection is visually indicated by changing the icon display of the slider knob.

In the case where the detection processing is detecting fast-forward, FIG.
(A) is displayed, and the display of the slider knob 603 is a fast-forwarding icon consisting of two triangles. In this case, a fast-forward video is displayed as the input image 601. Conversely, when the detection process is detecting details, FIG.
Is displayed, and the display of the slider knob 613 is a normal speed reproduction icon composed of one triangle. In this case, a normal speed image is displayed as the input image 601.

In the detection status display example shown in FIG. 6, the current detection processing position is further adjusted by using a slider (60 during fast forward).
3 and the others are visually represented by the position 613).
In this case, when the slider is at the left end, the head of the video is displayed, when the slider is at the right end, the end of the video is detected, and between the sliders, the detection process is performed for the video position corresponding to the ratio of the knob position.

In order to realize the detection status display method shown in FIG. 6, steps 415 and 432 may be modified as follows.

"Step 41 modified for the example of FIG. 6"
In 5 ”, the frame of aFrameFF is displayed, and then, FIG.
The icon of the slider 309 of FIG. 4 is changed to the knob 603 of FIG. 6A, and then the slider 309 of FIG. 4 is moved from the left end to (the position of the current frame image aFrameFF ÷ the total length of the input video). × the overall length of the slider,
To the position of.

"Step 43 modified for the example of FIG. 6"
In "2", the frame of aFrame is displayed, and then the icon of the slider knob 309 of FIG. 4 is changed to the knob 613 of FIG. 6B, and thereafter, the slider knob 3 of FIG.
09 from the left end at the position of (the position of the current frame image aFrameathe length of the entire input video) × the entire length of the slider.

FIG. 7 shows a visual display example of an event point whose detection position is ambiguous. FIG. 7 shows only the left part of the execution screen shown in FIG. 4 for explanation. The event point whose detection position is ambiguous is an event point detected when the detailed detection mode check box 307 in FIG. 4 is inactive, and this event point does not have a position accuracy of one frame unit. It is an imperfect event point. In the example of FIG. 7, the icon display when an event point whose detection position is ambiguous is visually distinguishable from the icon display of an event point whose detection position is not ambiguous.

7 is the same display area as the area 302 in FIG. In the notation 701, the icon of the event point whose detection position is ambiguous is displayed by multiplexing a transparent square with a dotted line before and after the icon and additionally displaying the characters “±?” On the display of the frame position. An accurate event point icon 703 can be visually distinguished. Thereby, the user can grasp at a glance which event point is an event point with one frame accuracy.

In order to realize a visual display example of an event point whose detection position is ambiguous in FIG.
The processing at the time of detection described in the above description may be modified as follows.

In the “detection process modified for the example of FIG. 7”, the special detection process includes a frame image of the event point at the end position of the icon list displayed in the area 302 of FIG. And a character string of a frame number representing the frame position of the event point in icon form
When the detailed detection mode described in step 414 is not established, a transparent dotted square like 701 in FIG. 7 is multiplexed and displayed before and after the icon image drawn earlier. , "±?" Is additionally displayed immediately to the right of the notation of the frame position number.

[0087]

According to the present invention, the following effects can be obtained. First, since the stream output at the time of detecting the fast-forward and the stream output at the time of detecting the detail are switched at any time according to the detection state of the event point, that is, the content of the stream, an output that does not disturb the reproduction order of the stream is provided. As a result, it is possible to easily confirm which part of the input stream is currently performing the detection processing.

Second, even if the user terminates the detection processing, the method of the present invention detects the event point position with one elementary data accuracy in each event point detection processing. Even if this is done, the event point position with one elementary data precision can be left.

Third, since the elementary data string required for the detail detection processing is always stored in the prefetch buffer,
Event points can be detected even for an input stream such as stream-format video data received as a TV broadcast, where data disappears immediately after output.

Fourth, the prefetch buffer and the event point detection method can be changed to a video or audio stream, and the event point detection method can be changed according to the detected event point. Event types can also be efficiently detected with a small amount of computation for types such as dissolve points and speaker change points of voice.

Fifth, since it is possible to easily confirm which portion of the input stream is currently performing the detection processing, an event point detection method that can be easily controlled by the user can be added.

[Brief description of the drawings]

FIG. 1 is a flow chart showing the outline of the method of the present invention.

FIG. 2 is a system configuration diagram of a stream event point detection display device using the method of the present invention.

FIG. 3 is a system configuration diagram for detecting a scene change point, which is a type of event point, from an MPEG file.

FIG. 4 is a diagram illustrating an execution screen in the implementation example of FIG. 3;

FIG. 5 is a processing flow diagram in the implementation example of FIG. 3;

FIG. 6 is a diagram illustrating an example of a method for visually displaying a detection status using a detection position slider.

FIG. 7 is a diagram illustrating an example of a visual display of an event point whose detection position is ambiguous.

[Explanation of symbols]

 1001 stream input device 1002 prefetch buffer 1003 fast-forward output control device 1004 detailed output control device 1005 fast-forward event point detection device 1006 detailed event point detection device 1007 detection control device 1008 stream output device

Claims (13)

[Claims] 1. A stream event point detection display method for detecting and outputting an event point such as a scene change point of a video or a speaker change point of a voice from input stream data of a video or an audio. The head section of the stream is input to the look-ahead buffer. (2) It is detected whether or not the event point exists in the look-ahead buffer. (3) If it is determined in (2) that the event point exists, the look-ahead is performed. While reading the raw data of the stream in the buffer (the minimum unit constituting the stream) sequentially and outputting the raw data to the outside, when the position of the raw data is determined as an event point, a detection process is performed. (4) (2) If it is determined that there is no event point, the stream in the prefetch buffer is fast-forwarded and output. (5) These (1) to (4) are sequentially repeated. More, and carrying out while suitably switched according to and outputs at the output or normal speed event point detection processing of the stream at fast forward stream content, the stream event point detection display method. 2. The stream event point detecting and displaying method according to claim 1, wherein the input stream is a video stream in which raw data is configured as a frame image, and the data in the prefetch buffer is composed of a frame image including several frames. By determining a video event point based on a feature amount such as an image difference of a frame image, the presence or absence of an event point can be detected in (2) of claim 1, and the frame position where the event point exists can be detected in (3). As a method of outputting raw data used in (3) and (4) of claim 1,
A stream event point detection display method, characterized by displaying a frame image or a frame image sequence on a display means such as a display device. 3. The stream event point detecting and displaying method according to claim 1, wherein the input stream is an audio stream whose raw data is composed of audio waveform data, and the data of the prefetch buffer is composed of a plurality of raw data concatenated. In the first aspect of the present invention, the presence or absence of an event point is determined by using a waveform feature amount such as the amplitude and frequency spectrum of the voice waveform. Time position can be detected, claim 1
A stream event point detection and display method characterized by outputting audio waveform data by an audio output means such as a speaker device as an output method of raw data used in (3) and (4). 4. The stream event point detection and display method according to claim 1, wherein in the event point detection method according to claim 1 (2), one elementary data in the prefetch buffer and a previous data read into the prefetch buffer are used. A stream event point detection and display method, wherein the presence or absence of an event point with respect to a stream in the prefetch buffer can be detected by obtaining a data feature amount with one elementary data in the data at that time. 5. The stream event point detection and display method according to claim 1, wherein the prefetch buffer does not read all of the elementary data strings constituting the stream section to be read, but the same as the head position of the prefetch section of the input stream. 3. The stream event point detection display according to claim 1, wherein the end position is stored, and the input stream is read only when the prefetch buffer is referred to in the event point detection method according to (1) or (2). Method. 6. The stream event point detection and display method according to claim 1, wherein the input stream is key element data which can be read in a short time, and non-key element data which takes a longer time to read than key data. In the case of the stream format composed of the following, the key reading data and the information described in claim 5 are recorded in the look-ahead buffer, and the event check described in claim 4 is performed as the event point detection device according to claim 1 (2). A method for detecting and displaying a stream event point, comprising using an output device. 7. The stream event point detecting and displaying method according to claim 1, wherein the stream event can be forcibly detected by a user during a detection process at a stream position currently being detected by a button or the like. Point detection display method. 8. The stream event point detection and display method according to claim 1, wherein a stream in which a user can ignore a process of detecting the position of an event point with one elementary data accuracy by a button or the like during the detection process. Event point detection display method. 9. The stream event point detecting and displaying method according to claim 1, wherein one detected event point is
A stream event point detection display method which can be displayed on a display device such as a display by a visual icon and an event point detection position. 10. A stream event point detection and display method according to claim 1, wherein the stream position currently being detected is displayed on a display device such as a display at any time. 11. The stream event point detection and display method according to claim 1, wherein whether the current detection processing status is a detection processing for each elementary data as described in (1) of claim 1 is performed. A stream event point detection display method in which the detection state can be visually distinguished by displaying an icon or the like at any time as to whether the detection processing is being performed in the fast forward as shown in (2) or (4). 12. The stream event point detecting and displaying method according to claim 8 or claim 9, wherein the user ignores the process of detecting the position of the event point with one elementary data accuracy by using a button or the like during the detection process. A stream event that can be displayed so that the display of an event point when detected in a state where it is detected and the display of an event point when detected in other states can be visually distinguished. Point detection display method. 13. A stream input device for reading an input stream, a look-ahead buffer for storing a head section of the read input stream, a stream output device for outputting a stream to a user or the like, A fast-forward output control device for outputting a stream in a fast-forward format, a detailed output control device for outputting a stream in the look-ahead buffer one by one, and a fast-forward for determining the presence or absence of an event point in the stream in the look-ahead buffer The event point detecting device, the detailed event point detecting device that determines the position of the event point in the stream in the prefetch buffer with one elementary data accuracy, and the fast-forward event point detecting device determines that there is no event point in the prefetch buffer. In this case, the contents of the pre-read buffer are output at the fast-forward output control device. If the fast-forward event point detecting device determines that there is an event point in the prefetch buffer, it sequentially reads the raw data in the prefetch buffer, displays the raw data on the detailed output control device, and displays the detailed event. A stream event point detection and display device, comprising: a point detection device for determining whether or not an event point exists at the position of the elementary data by the point detection device.