WO2011073559A1

WO2011073559A1 - Device and method for determining an acceptability score for zapping time

Info

Publication number: WO2011073559A1
Application number: PCT/FR2010/052697
Authority: WO
Inventors: Emmanuel Wyckens; Freddy Ceranton
Original assignee: France Telecom
Priority date: 2009-12-18
Filing date: 2010-12-13
Publication date: 2011-06-23

Abstract

The invention relates to a method for determining a zapping time acceptability score (Q) by measuring (E201) the zapping time between a first audiovisual scene and a second audiovisual scene following a zapping action (E202) in such a way as to also take into account at least one piece of information regarding the waiting content displayed between the first and second audiovisual scenes. The invention also relates to a parametric zapping time acceptability model for use in a determination method such as the one described. The invention further relates to a device implementing such a method and to a decoder including such a device.

Description

Device and method for determining a note of acceptability of the zapping time

The present invention relates to a method for determining a rating of acceptability of zapping time and quality of service measurement (QoS) associated with this zapping time.

More particularly, the invention is concerned here with a measurement of perceived quality or quality of experience (QoE) by the users on the zapping action.

The invention also relates to a device for determining an acceptability score capable of measuring the acceptability of the zapping time in audio / video devices.

Such determination or control makes it possible to evaluate the various audio / video systems such as set-top boxes or other broadcasters of audiovisual content.

The invention also relates to an innovative parametric model for use in such a method of determining a note of acceptability of the zapping time.

In existing systems for acceptability score determination, a measurement of the time between the zapping action to switch from a first audiovisual scene to a second audiovisual scene is performed. Depending on the zapping time thus measured, different parametric models have been defined in order to determine the level of user acceptability with respect to this time.

The document entitled "Perceived Quality of Channel Zapping" by Kooij, R., Ahmed, K., and Brunnstro, K., Proc. of 5th International Conference on Communication Systems and Network, Aug. 28-30, 2006, (2006), describes such a control system based on the measurement of the passage time between two chains.

An average score based on a quality scale called MOSp of acceptability scores called MOSp (for "Mean Opinion Score Prediction") comes from the models so applied.

However, the quality scores thus defined, do not exactly reflect the acceptability of the user to different situations that can be proposed between the first and second display screens.

Thus, it has been observed experimentally that a predetermined zapping time was more easily accepted by a user if a waiting content was displayed between the first and the second audiovisual scene.

There is therefore a need to take into account these differences in user perceptions to obtain a level of acceptability actually perceived by the user and therefore more reliable.

The invention improves the situation.

For this purpose, the invention relates to a method for determining a rating of acceptability of a zapping time from a measurement of the zapping time between a first audiovisual scene and a second audiovisual scene following a zapping action. The method is such that it comprises the following steps:

- Detecting information of the waiting content displayed between the first audiovisual scene and the second audiovisual scene;

selecting a parametric model from a plurality of stored parametric models according to the detected information and the measured time;

determination of an acceptability score from the selected parametric model. Depending on the information detected in the waiting content, the parametric models differ. Selecting the right parametric model will give the acceptability score appropriate to the situation. This acceptability note also takes into account at least one information on the waiting content displayed between the first and the second audiovisual scene.

Thus, the acceptability score is determined by taking into account information on the waiting content that can vary this rating with respect to a determination based solely on the zapping time.

This note is therefore more representative of the real feeling of the user and gives a real measure of quality of experience (QoE).

The various particular embodiments mentioned below may be added independently or in combination with each other, with the characteristics mentioned above.

Thus, in particular embodiments, the detection of displayed information comprises one or more detection steps of the following group:

- fixed image detection on all or part of the display,

- black image detection,

- frozen image detection of the first or second audiovisual scene,

detection of a banner of text information,

- motion picture detection,

- detection of absence of sound.

These different detection methods thus make it possible to analyze the waiting content and to detect either black-image content, still or frozen image, animated image for example a "chaser", a sound cut or a sequence of several types mentioned. The possible detection list is absolutely not exhaustive and can include many other types of detections adapted to various waiting contents.

In a particular embodiment, it further comprises a measurement of the display times between different detected information, the measured times being used for the selection of the parametric model.

It is thus possible to differentiate also waiting contents according to display sequences of different contents and the display time of these contents. Adapted parametric models are thus constructed according to these characteristics and stored for control use.

In a preferred embodiment, the parametric models are determined by learning, from zapping time measurements and information on the waiting contents.

The parametric models are therefore adapted to the waiting contents that can be displayed.

The invention also relates to a parametric model of acceptability of the zapping time intended to be used in a method of determining a zapping time acceptability score, determined from parameters for measuring the zapping time between a first audiovisual scene and a second audiovisual scene. This model is such that it is further determined from information on a waiting content displayed between the first and the second scene.

This type of parametric model, unlike existing models, takes into account wait-content information. Different parametric models can thus be determined and stored for use in a determination method as described above. In a particular embodiment, the parametric models take into account it is further determined from measured display times between different displayed information of the waiting content.

The invention relates to a device for determining a note of acceptability of the zapping time comprising a module for measuring the zapping time between a first audiovisual scene and a second audiovisual scene following a zapping action. The device is such that it further comprises a module for obtaining at least one information on the waiting content displayed between the first and the second audiovisual scene for determining the acceptability score of the zapping time.

It also relates to an audio / video decoder comprising a device as described above.

Such a decoder can be integrated in a lounge decoder, a "set top box", a video player or any device for reading or broadcasting audio / video content.

Finally, the invention relates to a computer program comprising code instructions for implementing the steps of a determination method as described, when these are executed by a processor. Other features and advantages of the invention will emerge more clearly on reading the following description, given solely by way of non-limiting example, and with reference to the appended drawings, in which:

FIG. 1 illustrates a system for determining a note of acceptability of the zapping time comprising a device for determining a note of acceptability of the zapping time according to one embodiment of the invention; FIG. 2 illustrates a method for determining a note of acceptability of the zapping time according to one embodiment of the invention;

FIG. 3 illustrates an example of a sequence of audiovisual scenes and waiting contents, displayed on a display screen;

FIGS. 4a to 4f illustrate several examples of waiting contents displayed during the zapping time;

FIG. 5 illustrates the steps of detection of reference sequences in a waiting content according to a particular embodiment of the invention;

FIG. 6 illustrates the steps of sound detection of a waiting content according to a particular embodiment of the invention;

FIGS. 7a and 7b illustrate examples of parametric models according to the invention; and

FIG. 8 illustrates an embodiment of a device for determining a score of acceptability according to the invention.

FIG. 1 illustrates a system for checking the acceptability of the zapping time in which is represented a screen for viewing audiovisual content 170 such as a television, an audio / video decoder 160 or "set top box" or any other reader or broadcaster of audiovisual content.

Such a decoder comprises a video acquisition module 161, an audio acquisition module 162, a module 163 for detecting infrared commands coming for example from a remote control 180 and a control module 164. These modules are conventional and well known. and will not be more detailed here.

The decoder is connected to a device 100 for determining a note of acceptability of the zapping time. In another embodiment, this device can be an integral part of the decoder. The determination device 100 receives as input a video stream and an audio stream. These flows are analyzed by a module of detection 110 which is able to detect different information that can be displayed as waiting content during a change of audiovisual scenes following a zapping action. This zapping action is detected by the module 120 for filtering infrared codes.

Several detection modules are integrated in the module 110. Here is illustrated a reference image detection module 111 (a reference image being for example a black image), a fluid loss detection module 112 which may for example show that an image is frozen, an information text banner detection module 113, a logo or moving image detection module 114 and a sound stop detection module 115.

Other detection modules can thus be provided according to the different waiting contents that can be displayed via the decoder.

At the end of these detection modules, the information thus detected is timestamped by the time stamping module 130. Thus different times are measured according to the display sequences of the different information of a waiting content.

A module for simultaneity detection of audio and video renditions is also illustrated at 150. This detection also makes it possible to check how the waiting sequences are decoded. The level of acceptability for the user of these different sequences varies. The simultaneity module of the audio and video renderings verifies the synchronism of the waiting contents as well as the sequence of a change of audiovisual scenes. During the wait or the appearance of the new program, the broadcasting of the audio signal before the display of the video can improve the perception of the zapping time, and the opposite can penalize this same perception.

All the information on the waiting content as well as the zapping time measured at 130, are given as input of a selection module 140 which will selecting a parametric model from among a plurality of stored patterns, which corresponds to the detected content and the measured zapping time.

The particularity of the parametric models as stored is that they are determined not only according to a measurement parameter of the zapping time but also from information on the waiting content.

The parametric model itself is original.

The determination of this type of model is done by learning after a series of subjective tests done among a panel of users who give acceptability notes according to the waiting contents they visualize and the zapping times they undergo. .

Examples of such parametric models will be illustrated later with reference to FIGS. 7a and 7b.

The application of this parametric model selected at 140 to the detected and measured information makes it possible to obtain a Q acceptability score at the output of the device.

This acceptability score can then be transmitted to a processing server which analyzes this note to evaluate the decoder thus tested, and thus detect, for example, a potential defect on the broadcasting of the content, on the decoder itself in relation to a note. acceptability that the server would likely expect.

In a possible use of such a device 100, it can be placed at a specific location of the production network to test several different decoders and evaluate these different decoders. This device then has a role of evaluation robot.

With reference to FIG. 2, the steps of a method for determining a note of acceptability of the zapping time are now described. A first step of the E200 control method is to detect a zapping action (Detec.tzap). At the detection of this zapping action to switch from a first audiovisual scene to a second audiovisual scene, a stopwatch is started at step E201 to measure (Horodat.) The total time of passage from one scene to another .

The zapping time is in fact the addition of network control times for changing the channel or content, storing the audio / video information and the decoding time of the second audiovisual scene.

Meanwhile, a step E202 detection (Detectlnf.) Information displayed between the first scene and the second scene is performed.

Several types of detections can be made depending on the type of waiting content that can be displayed. Some examples of detections will be described later with reference to FIGS. 5 and 6.

An optional step E203 for determining the information display time of a waiting content sequence can also be performed.

When the second audiovisual scene is displayed (At.Sc2) in E204, the total zapping time (t.zap.) Is determined at E 205.

This zapping time information and the information from the detection on the waiting content are used in step E 206 to select a parametric model (Param.Model) adapted from among a plurality of stored parametric models.

The application of this selected parametric model to the information obtained in step E205 and step E202 and optionally in step E203, makes it possible to determine (Det.Q) an acceptability note Q in step E207. This method thus determines a note of acceptability of the zapping time from the measured zapping time and taking into account information on the waiting content displayed during this zapping time.

This determination brings a real vision of the quality of experience (QoE) of the user.

Figure 3 illustrates an example of a sequence of audiovisual scenes and a waiting content following a zapping action between two audiovisual channels. After a zapping action, represented here at 301 at the instant T_C_C_1, the passage of a channel A displaying a first audiovisual scene (Sel) to a scene B displaying a second audiovisual scene (Sc.2) is effected by a display of a waiting content (Cont). In the example represented here, the waiting content is composed of an arimixed image 302 in the form of a "chaser" and a 303 textual information (EPG) strip for "Electronic Program Guide". information on the progress of the current program.

This figure also illustrates the video acquisition that starts at a moment

T_L_A_video. The duration of the acquisition is represented in 304 and is here called D_A_video. During this duration of acquisition, the images 1 to n of the chain A (Im.l A, Im.2 A, ..., Im.n A) are displayed for scene 1 of the chain A, the images 1 to n (Im.l Cont., ..., Im.n Cont.) of the waiting contents, the images m and m + 1 (hn.rn B, ..., Im.m + l B) of stage 2 of the B chain.

The video acquisition is preferably done on different video connections for example, on HDMI jacks (For "High Definition Multimedia Interface" in English) or scart jacks. Video signals can be scanned into YUV or RGB component formats (for Red, Green, Blue) to obtain video samples. These video samples can come directly from the output of a video decoder. The start and / or end of acquisition action video is initiated by a task scheduler or controller as shown at 164 in FIG.

Likewise, the audio acquisition is illustrated by a plurality of audio samples (Audio Echo 1A, Audio 2A Echo, Audio Echo n A) for the A string and (Audio m B, ... , AudioAudio m + 1 B) for the chain B. The audio acquisition time is represented at 305 and is called D_A_audio. It starts at the moment T_L_A_audio.

The audio acquisition is preferably done on the various audio connectors such as HDMI jacks or scart jacks. Audio signals can be digitized in PCM format (for "Pulse Code Modulation") to obtain audio samples. These samples can also come directly from the output of the video decoder. The start and / or end of audio acquisition action is also started by the task scheduler or controller 164. These actions can be started simultaneously with the actions on the video capture.

Chain change control can be by means of infrared code transmission or via decoder key management.

The instant of the launch of the zapping execution code is here represented at 301, at the instant T_C_C_1. H is greater than the time T_L_A audio and T_L_A_video.

In this example illustrated in FIG. 3, a waiting content comprising an animated "chase" and an information banner is displayed during the zapping time.

Figures 4a to 4f illustrate different waiting contents with one or more contents displayed. Thus, in FIG. 4a, the waiting content is a black image with a textual program information banner (EPG), the hour (h) and a cursor informing of the progress of the program in progress. . In FIG. 4b, the waiting content further comprises the information of FIG. 4a, an animated image in the form of a "chaser".

In FIG. 4c, the waiting content is a frozen image, that is to say a still image which generally corresponds to the last image displayed for the scene before the zapping action. In addition to this still image, the waiting content also includes the information banner described above.

In FIG. 4d, the waiting content further comprises the information described in FIG. 4c, an animated image in the form of a "chaser".

In FIG. 4e, the waiting content comprises a sequence where a first frozen image is displayed with the information banner, then a black image is displayed with this same banner.

In FIG. 4f, the same sequence is displayed with an animated image in the form of a "chase" for each of the images displayed.

Of course, these illustrations represent only possible embodiments. Many other waiting contents can be displayed with one or more images in the form of sequences, with images or logos, animated or not. The audio part of the waiting content can also intervene. For example, the sound of the first scene may remain during a time interval less than or equal to the zapping time or the sound may be cut during the entire duration of the zapping.

In order to detect these different waiting contents, a step of detecting information of the waiting content is performed. This step may contain one or more detection sub-steps. These sub-steps may for example be black image detection, still image detection, frozen image, textual or animated information strip detection, motion picture detection, stop detection. sound, etc ... Figure 5 illustrates one of these detections. A detection of reference images in the waiting content is thus illustrated. For a specific decoder or determined television service, the waiting content includes the same images or sequences of images. These images constitute reference images that can for example be recorded in this decoder after a training sequence via video acquisitions. This learning sequence is performed before the content analysis phase. Thus reference images Im.Irl to Im.Irn as illustrated in FIG. 5 at 501 are recorded.

The detection of these reference images in the waiting content is performed by a comparison between these reference images and the images of the waiting sequence displayed. This comparison can be done on all or part of the displayed image, in particular to avoid the area of appearance of the information banner. A tolerance level on the luminance and chrominance of the compared images is applied.

The following formula is for example applied to compare point by point the displayed image and the reference images.

Captured images (x, y, t) - Image Irn (x, y) - 0 (1) To temporally locate the position of the reference images in the captured and displayed waiting sequence, the following formula is applied:

TJPJJRn = T_L_Avideo + (1 / ps x PJmgn) (2) where T_P_I_Rn corresponds to the moment when the reference image "n" is found among the captured images, PJmgn corresponds to the position of the reference image "n "in the set of captured images, fps corresponds to the capture frequency (for example 50 or 60 Hz), TJ_A_video corresponds to the start time of the video acquisition. T_P_I_Rm is the time when the reference image "m" is found among the captured images. This image m corresponds to the last reference image found.

Assuming that one or more images ("Im" to "Im.Irn") usually exist in the sequence of waiting contents and that the reference images are not all found over the analysis period of the captured sequence, we can consider that there are no reference images in the captured sequence. In this case, the note of acceptability of the zapping time can not be determined.

In the illustration of FIG. 5, the reference images are compared with the images captured at 502 and 503 to define the images present in the waiting content between the instants T_P_I_Rn and T_P_I_Rm.

In the case where a reference image detection is performed (output O at 503), this information is kept in memory to select the right parametric model in step 206 of FIG.

In the case where no reference image is detected in the waiting content or the normally expected sequence is not found again (output N in 503), this means that the waiting content is not consisting of these images. At this stage, it is not possible to choose a parametric model adapted to the waiting content. Another detection of information (Detect + 1) on this waiting content will then be implemented to find its composition and select the right parametric model.

Another possible detection on the waiting content is that of the loss of fluidity of the video stream. This detection is intended to verify whether the video sequence is performed on one or more stable or stable image rate. The fluidity detection must take place before the change of control command chain. The detection is started before the instant T_C_C_1 to see if there is a loss of video fluidity on the existing program, until the arrival of the new program.

The scan time of the video detection may be equal to D_A_Video. The loss of fluency detection will be performed on the images of the current channel and on the images of the new requested channel. In order to predict a possible loss of fluidity, the number of images of the current captured string must be greater than or equal to the number of images over a period of one second (ie the inverse of the image frequency). For example, if the frame rate is 25 frames per second, you must capture at least 25 frames on the current channel.

The reasoning is identical for the images of the chain B. The images of the chains A and B can sometimes be interspersed with images of waiting with or without the presence of an animated logo, a fixed logo or related information. in the name of the channel and the associated program. This search is carried out by the detection of reference images or by the detection of fixed or animated logos.

The loss of fluidity is therefore related to the presence of a momentary gel. The image freeze is detected by calculating whether the derivative of the luminance and chrominance levels with respect to time is zero, according to the following equations:

StringA (x, y, t) - String A (x, y, t + I) = 0. (3)

StringB (x, y, t) - Image string B (x, y, t + I) = 0. (4)

Another example of detecting information on the waiting content is illustrated with reference to FIG. 6. In this figure, the audio samples of the channel A are illustrated from the start of the audio acquisition T_L_A_audio. The instant of start of zapping is represented in this figure by T_P_I_Am. From this moment, the samples of sound waiting are represented in ContAudio 1 and ContAudio n.

When the zapping time is over, the audio samples of the B channel are available.

To measure the sound level loss of the sound samples or the mute of the waiting content, a measurement of comparison to an audio threshold (Sax) is performed on N sound samples. It can indeed be considered that there is a mute when the detection of N samples less than a threshold is greater than a predetermined time, for example 30ms. This time also corresponds to a number of samples M.

The following formula is then applied to detect the instant of non-presence of the sound and therefore the beginning of the zapping time:

T_P_I_ARn = T_L_A_a dio + Mx (l / audio sampling frequency) (5) where M is the number of samples after which the sound is stopped. T_P_I_ARm corresponds to the time when the sound level returns above the Sax audio threshold, that is to say that N samples greater than the threshold are detected over a period greater than a predetermined duration, for example 30ms.

For an information banner detection of the waiting content, a method as described with reference to FIG. 5 may be used with a spatial location on an area of the image.

One can also detect the appearance of logos or moving images on the waiting content by locating the comparison of reference images to a specific area of the display. It is thus possible to define the instants of appearance of the logo and the disappearance of this same logo. With this information on the audio and video content of the waiting content, it is possible to choose the parametric models adapted to this type of content and to the measured waiting time.

For this, the parametric models are determined by learning from a series of subjective tests carried out with a panel of users testing different waiting contents and different zapping times.

Thus, the parametric models are defined both as a function of the zapping time between two audiovisual scenes but also by taking into account information on the waiting content displayed between these two scenes.

In a general way, a parametric model taking into account these two elements is of the following form:

Q = 100 / (1 + (a DZ) ^-b ) (6)

With Q, the acceptability score, DZ, the zapping time measured and a and b, the coefficients to be determined according to the waiting contents.

The coefficient a is for example between 1000 and 4000 and the coefficient b is between 4 and 10.

Particular examples of parametric models, with coefficients a and b determined according to the waiting content, are illustrated with reference to FIGS. 7a and 7b.

In the case where the waiting content is a black image without any video sequence and the sound is not present, the end times of detection of the reference images (here, black image) (T_P_I_Rm in FIG. 5) and the end times of detection of absence of sound {T_P_I_Arm in Figure 6), are substantially equivalent.

The duration of zapping is then determined as follows:

DZ = T_PJ_Rm - T_C_C_1 (7) The acceptability of the zapping time for this type of waiting content is defined by the following parametric model:

Here Q represents the acceptability score of the zapping time. It is expressed as a percentage.

FIG. 7a and FIG. 7b graphically illustrate this parametric model as a function of the measured zapping time.

In the case where the waiting content is a frozen image without any video link and the sound is not present, the zapping duration is determined as in equation (7). The end of detection of the frozen image corresponding to the moment when the equations (3) and (4) are no longer verified.

The acceptability of the zapping time for this type of waiting content is defined by the following parametric model:

Figure 7b graphically illustrates this parametric model, as a function of measured zapping time.

In the case where the waiting content is an animated logo displayed for example on a black image and the sound is not present, the end times of detection of the reference images (here, black image) (T_P_I_Rm on the FIG. 5) and the end of detection moments of absence of sound (T_PJ._Arm in FIG. 6) are substantially equivalent. The moments of detection of the beginning of display of the animated logo and the end of display are in fact between these two moments of beginning of detection of the black image T_P_I_Rn and the end of detection of the black image T_P_I_Rm, The duration of zapping is then determined in the same way as the equation

(7).

Figure 7a graphically illustrates this parametric model as a function of measured zapping time.

These parametric models are non-exhaustive examples of models that characterize waiting contents. Many models can thus be determined for content containing only one type of information displayed as for content that includes sequences of several information displayed.

In the same way, the models taken as an example here are determined according to the total time of zapping. Other models may also be expressed as a function of intermediate times of display of the different information of the waiting content.

FIG. 8 represents an example of a device for determining an acceptability score. This device comprises a PROC processor cooperating with a memory block BM having a memory storage and / or working MEM.

The memory block may advantageously comprise a computer program comprising code instructions for implementing the steps of the method of determining a score of acceptability in the sense of the invention, when these instructions are executed by the processor PROC, and in particular measuring the zapping time between a first audiovisual scene and a second audiovisual scene following a zapping action and taking into account at least one information on the waiting content displayed between the first and the second audiovisual scene. Typically, the description of FIG. 2 repeats the steps of an algorithm of such a computer program. The computer program may also be stored on a memory medium readable by a reader of the device or downloadable in the memory space of this device.

The device comprises an input module (E) able to receive an audio signal

(A) and video (V) as well as zapping control information (Zap.). This information comes either from a decoder or external reader, or from an internal module when the device is integrated in such a decoder or reader.

The device comprises an output module (S) capable of transmitting an acceptability note Q to a communication network or other information transmission device.

Claims

A method for determining an acceptability score (Q) of a zapping time from a measurement (E201) of the zapping time between a first audiovisual scene (Sc.l) and a second audiovisual scene ( Sc.2) following a zapping action (E200), characterized in that it comprises the following steps:

- detecting (E202) information of the waiting content displayed between the first audiovisual scene and the second audiovisual scene;

parametric model selection (E206) among a plurality of parametric models stored according to the detected information and the measured time;

determination (E207) of an acceptability score from the selected parametric model.

2. Method according to claim 1, characterized in that the detection of displayed information comprises one or more detection steps of the following group:

- fixed image detection on all or part of the display,

- black image detection,

- frozen image detection of the first or second audiovisual scene,

detection of a banner of text information,

- motion picture detection,

- detection of absence of sound.

3. Method according to claim 1, characterized in that it further comprises a measurement of the display times between different detected information, the measured times being used for the selection of the parametric model.

4. Method according to claim 1, characterized in that the parametric models are determined by learning from measurements of zapping time and information on the waiting contents.

5. Device for determining a note of acceptability of the zapping time comprising a module for measuring the zapping time between a first audiovisual scene and a second audiovisual scene following a zapping action, characterized in that it comprises in in addition to a module for obtaining at least one information on the waiting content displayed between the first and the second audiovisual scene for determining the acceptability score of the zapping time.

6 Audio / video decoder comprising a device according to claim 5.

Computer program comprising code instructions for the implementation of a checking method according to one of claims 1 to 4, when these are executed by a processor.