JP2012168780A - Change responsive preference estimation device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To perform preference estimation in consideration of individual differences and changes in preference caused by differences in audiovisual characteristics and feelings of users when viewing content.SOLUTION: A change responsive preference estimation device includes a feature amount extraction unit 12 for extracting the features of content, a resolution application unit 12a for applying a filter corresponding to a resolution for each user in the feature amount extraction unit 12, and a resolution estimation unit 30 for changing the filter values of the resolution application unit 12a according to content viewing experience of a user.

Description

  The present invention relates to a preference estimation technique for estimating a user's preference for content in content search, authoring, recommendation, and the like.

  Conventionally, as a system for estimating a target user's preference from content features, the content feature amount is analyzed, and the association between the feature amount and the user's preference degree for the content obtained through a questionnaire or the like is learned. There is an object that estimates the preference for the new content by analyzing the feature amount of the new content. For example, when music content is targeted, the preference is estimated by using beats, tonality, melody structure, and the like representing each music content as feature amounts of the content. For example, for beat-related features, the tempo representing the speed of the song, the beat structure indicating the basic beat structure such as 8 beats and 16 beats, the beat strength of 1/2 beat and the beat of 1/4 beat There is a beat intensity ratio representing an intensity ratio, an average number of sounds, and the like. Tonality refers to the key tones of music such as C major and D minor. As a technique for expressing the melody structure, there is GTTM for summarizing music. In GTTM, summarization is performed by selecting the most important sounds in units of bars or in units of several bars and creating a tree structure. In addition, vocal voice quality and musical instrument composition are also feature quantities.

  A representative method for estimating user preferences from music content features is Pandora Media, Inc. Music recommendation system (www.pandora.com, Music Genome Project [Patent Document 1]).

  The configuration of a conventional preference estimation device is shown in FIG. This preference estimation apparatus includes a feature amount extraction unit 12 that extracts feature amounts of input content, and a preference estimation unit 11 that estimates preferences from the similarity of the feature amount to the feature amount preferred by the target user. Here, when the content is input, the feature amount extraction unit 12 calculates the feature amount and inputs the feature amount to the preference estimation unit 11. The preference estimation unit 11 determines a user's preference based on the input feature amount, and outputs a preference determination result.

US7000315 JP 2006-195384 A

Tagawa and Misaki: “A study on tempo detection from music signals”, Acoustical Society of Japan, pp. 529-530, 2000 Rhythm map: Kakuo, Ono, Kashiyama: “Extraction of unit rhythm pattern from music acoustic signal and analysis of music structure”, Information Processing Society of Japan, Music Information Science, 2008-MUS-76, p149-154

  However, in the conventional preference estimation device, each user does not handle in consideration of which part of the extracted content feature quantity, for example, voice quality, instrument configuration, beat structure, etc. For this reason, for example, for a music piece selected with a preference for the beat structure, only the female vocal included in the music is focused, and a similar female vocal music is recommended. There was a case. In addition, over time, the user's preference for listening to the content has changed, and the preference has changed because no consideration is given to changes in the preference, such as the fact that the user likes previously becomes boring, or on the contrary Even so, there was a problem of recommending the same content forever.

  The present invention has been made in view of the above-mentioned reasons, and the preference estimation is performed in consideration of individual differences and changes for each user in the audiovisual ability for viewing contents in detail and the resolution representing the fineness of the way of feeling. An object is to provide a possible preference estimation device.

  In order to solve the above-described conventional problem, the change correspondence type preference estimation apparatus of the present invention includes a feature amount extraction unit that extracts content features, and a resolution that applies a filter corresponding to the resolution for each user in the feature amount extraction unit. An application unit and a resolution estimation unit for changing the filter value of the resolution application unit according to the user's content appreciation experience are provided.

According to this configuration, it is possible to estimate the preference of individual individuals at a certain point in time and slightly ahead, taking into account differences and changes in content preference for each user.
In addition, the change-response-type preference estimation device according to the present invention further includes a caution estimation unit for changing a change characteristic of the filter value of the resolution estimation unit according to how the user pays attention to the content. It may be a thing.

  According to this configuration, the feature amount can be extracted in consideration of the viewer's attention level with respect to each content. Therefore, it is possible to grasp the change in the preference of the content more accurately when extracting the feature amount.

  In the change correspondence type preference estimation apparatus according to the present invention, the resolution estimation unit may be a frequency characteristic estimation unit that estimates the frequency resolution of the user's auditory characteristic, or all or any one of the rhythm characteristic estimation units that estimate the rhythm resolution. It may have more than one.

According to this configuration, it is possible to grasp changes in content preference more accurately when extracting feature amounts for each content.
In the change correspondence type preference estimation apparatus according to the present invention, the resolution estimation unit further includes a feature amount of the content after being filtered by the resolution application unit viewed by the user in the past, at the time of listening estimated by the attention estimation unit. There may be provided a listening history management unit for managing the listening history including the attention information.

  According to this configuration, it is possible to perform preference estimation in consideration of the feature amounts of a large amount of content by managing the feature amounts and the like of a large amount of content that has been listened to in the past by the listening history management unit.

  The present invention is also a change-aware preference estimation method in a change-aware preference estimation device that estimates a user's preference for content, and includes a feature quantity extraction step for extracting content features, and a user in a feature quantity extraction unit. A change application type preference estimation method including a resolution application step for applying a filter corresponding to the resolution of the image, and a resolution estimation step for changing the filter value of the resolution application unit according to the user's content viewing experience. Also good.

According to this configuration, it is possible to estimate the preference of individual individuals at a certain point in time and slightly ahead, taking into account differences and changes in content preference for each user.
The present invention is also a change-aware preference estimation program for causing a computer to function as a change-aware preference estimation device that estimates a user's preference for content, wherein the computer extracts feature amounts for extracting content features. A resolution application step for applying a filter corresponding to the resolution for each user in the feature amount extraction unit, and a resolution estimation step for changing the filter value of the resolution application unit in accordance with the user's content appreciation experience. It may be a change correspondence type preference estimation program to be executed.

According to this configuration, it is possible to estimate the preference of individual individuals at a certain point in time and slightly ahead, taking into account differences and changes in content preference for each user.
In addition, the present invention provides a feature amount extraction unit that extracts content features, a resolution application unit that applies a filter corresponding to a resolution for each user in the feature amount extraction unit, and a filter value of the resolution application unit as a user value. It may be a change correspondence type preference estimation integrated circuit including a resolution estimation unit for changing according to the content appreciation experience.

  According to this configuration, it is possible to reduce the size of the entire apparatus including the content recommendation integrated circuit.

It is a figure which shows the whole structure of the change corresponding | compatible preference estimation apparatus which concerns on embodiment. It is a figure which shows the structure of the feature-value extraction part which concerns on embodiment. It is a figure which shows the structure of the resolution estimation part which concerns on embodiment. It is a figure which shows an example of the resolution / history management table managed by the resolution / history management unit 33 according to the embodiment. It is a figure which shows the listening experience value / frequency characteristic correlation table which concerns on embodiment. It is a figure which shows the listening experience value / rhythm characteristic correlation table which concerns on embodiment. It is a figure which shows the structure of the attention estimation part which concerns on embodiment. It is a whole block diagram of the apparatus which performs preference estimation of a music content among the change corresponding | compatible type preference estimation apparatuses which concern on embodiment. It is a flowchart which shows operation | movement of the change corresponding | compatible preference estimation apparatus which concerns on embodiment. It is a flowchart which shows the rhythm pattern extraction process which concerns on embodiment. It is a flowchart which shows the feature-value extraction process which concerns on embodiment. It is a flowchart which shows the frequency characteristic estimation process which concerns on embodiment. It is a flowchart which shows the rhythm characteristic estimation process which concerns on embodiment. It is a flowchart which shows the management table update process which concerns on embodiment. It is a figure which shows the whole structure of a prior art example.

<Embodiment>
<1> Configuration <1-1> Overall Configuration FIG. 1 shows a configuration of a change-aware preference estimation apparatus according to the present embodiment.

The preference estimation unit 11 estimates the user's preference for the input content.
The feature amount extraction unit 12 is a feature amount that represents the feature of the content, such as a music feature amount such as beat or frequency characteristic in the music content, or a SIFT feature amount in the image content, from the signal representing the input content. Is extracted and output.

  The feature amount extraction unit 12 includes a resolution application unit 12a that sets a filter to be applied depending on a difference in appreciation power when a user views content. This appreciation power is the ability to discriminate between good and bad content. For example, if the content is music, it refers to the ability to understand the details of the composition, arrangement, performance, spatiality, etc. . The filter is, for example, a frequency band filter if the content is music, or a color filter if the content is music.

  The resolution estimation unit 30 calculates a current value (filter value information) of a filter with a resolution for each user set in the resolution application unit 12a. Then, the resolution estimation unit 30 inputs the calculated filter value information to the feature amount extraction unit 12. Here, the resolution means, for example, the ability to distinguish fine rhythms or hear sounds of various frequency bands, that is, the ability to support appreciation.

The attention estimation unit 22 estimates the attention information indicating what the user is watching the content carefully from the brain measurement information of the user or the information input by the user. Here, the attention information is, for example, in the case of music content, mainly about what the user is interested in, such as vocal voice quality or electric guitar technique, and what to focus on. This means information that indicates whether you are tilting. This attention information will be described in detail later.
<1-2> Feature Quantity Extraction Unit FIG. 2 shows the configuration of the feature quantity extraction unit 12.

The frequency spectrum calculation unit 12b calculates the frequency spectrum of the music signal of the input music content.
The feature amount extraction unit 12c extracts a music feature amount from the frequency spectrum obtained by performing filtering using the frequency filter associated with each user by the resolution application unit 12a. The feature amount extraction unit 12c includes a feature amount storage unit (not shown) that stores music feature amounts, and stores the extracted music feature amounts in the feature amount storage unit.

  In the resolution application unit 12a, an individual frequency characteristic FP and an individual rhythm characteristic RP are set as resolution filters (filter value information) associated with each user calculated by the resolution estimation unit 30.

  The value of the individual frequency characteristic FP is a value serving as an index representing the auditory characteristic of each user. As the FP value, an identification number of a resolution filter (resolution filter related to the frequency characteristic FP) can be used. A method for calculating the individual frequency characteristic RP will be described later.

  The value of the individual rhythm characteristic RP is a value that serves as an index of how much complicated rhythm the user can distinguish. As an RP value, an identification number of a resolution filter (resolution filter related to rhythm characteristics) is used, and a user who hears only simple notes such as 4 minutes or 8 minutes has a value close to “0”, 16 minutes or 32. A user who hears a complicated rhythm pattern using minute notes such as minute, triplet, and syncopation may be defined to take a value close to “1”. However, this value can be defined according to the system implementation. A method for calculating the individual rhythm characteristic RP will be described later.

In the following description, it is assumed that the FP value and the RP value are resolution filter identification numbers.
<1-3> Resolution Estimator Next, the configuration of the resolution estimator 30 will be described with reference to FIG.

  In FIG. 3, a frequency characteristic estimation unit 31 calculates a listening experience value FE for a frequency for each user, and estimates a frequency characteristic for each user based on the listening experience value FE. The rhythm characteristic estimation unit 32 calculates the listening experience value RE for the rhythm for each user, and estimates the rhythm characteristic for each user based on the listening experience value RE. The listening history management unit 33 manages the listening history for each user. The calculation method of the listening experience value FE with respect to the frequency and the calculation method of the listening experience value RE with respect to the rhythm will be described later.

  As shown in FIG. 4, the listening history management unit 33 includes a unique user ID for each user, individual frequency characteristics indicating the current music listening resolution of the user corresponding to the user ID, and information on individual rhythm characteristics. It manages a resolution / history management table composed of a user resolution table (upper part in FIG. 4) and a user listening history table (lower part in FIG. 4) comprising information relating to the history of each piece of music that has been listened to. Each time the user listens to the music, the user resolution table and the user listening history table are updated. As shown in the lower part of FIG. 4, information on the history of each listened music that the user has listened to is a structure having the music ID of the listened music and the music feature stored in the feature storage at the time of listening as members. Pointers (P [0], P [1],..., P [M] in FIG. 4), an individual frequency characteristic value FP and an individual rhythm characteristic value RP which are identification numbers of resolution filters at the time of listening. And attention information that is output from the attention estimation unit 22 and indicates the user's attention state to the music content at the time of listening, and the listening time of each listening music. The user listening history table is associated with the user ID. For example, when there are a plurality of user listening history tables, the user ID can be selected as a key.

In FIG. 4, the value of “user ID” is a unique number assigned to each user. In addition, the value of “individual frequency characteristic FP” in the “current resolution” column in FIG. 4 is a “No.” column indicating the identification number of the resolution filter in the listening experience value / frequency characteristic correlation table (see FIG. 5) described later. "1" means the lowest resolution, and "3" means the highest resolution. The value of “individual rhythm characteristic RP” in the “current resolution” column corresponds to the “No.” column indicating the identification number of the resolution filter in FIG. 6, “1” means the lowest resolution, and “3” "Means the highest resolution. In addition, the value of “Song ID” in the “History” column of FIG. 4 is a number unique to the listened music. In the user listening history table shown in the lower part of FIG. 4, even when a song having the same “song ID” (for example, a song having a song ID “1” in FIG. 4) is repeatedly listened to, it is managed as separate songs. ing. 5 stores the values of the individual frequency characteristics FP and the individual rhythm characteristics RP at the time of listening to each piece of music (the “frequency characteristics” field and the “rhythm” field in FIG. 4). Refer to the “Characteristics” column). The “attention information” column in FIG. 4 is a value obtained by the attention estimation unit 22 to be described later, and represents how carefully the user listened to music during listening. “A” means listening, “B” means attention to vocals, and “C” means attention to the composition of the whole music. Further, this “attention information” field is expressed by an integer value, and in this embodiment, A = 1, B = 2, and C = 3 are set. These numerical values are used as the attention information “a” in [Equation 1] and [Equation 2] described later. The format of the management table shown in FIG. 4 is an example and is not necessarily limited to this format.
<1-4> Frequency Characteristic Estimation Unit The frequency characteristic estimation unit 31 uses the information in the user listening history table of the resolution / history management table managed by the resolution / history management unit 33, and the relational expression of [Equation 1]. Is used to calculate a listening experience value FE, which is a value indicating the total listening experience related to the frequency (step S311).

ここにおいて、ｉは楽曲のＩＤ（曲ＩＤ）、ｓｆはその楽曲の周波数成分に関連する特徴量、ｔはその楽曲の聴取時間、ａはその楽曲を聴取した際の注意情報を示す。ここで、聴取時間は、その楽曲を実際に聴いた時間を指している。例えば、３分の楽曲のうち最初や途中の１分を聴いた場合、聴取時間は１分となる。また、聴取経験値ＦＥの算出に用いるｓｆは、曲ＩＤに対応付けされたポインタ（図４参照）により特定される。

Here, i is a song ID (song ID), sf is a feature quantity related to the frequency component of the song, t is the listening time of the song, and a is caution information when listening to the song. Here, the listening time indicates the time of actually listening to the music. For example, when listening to the first or middle one minute of a three-minute song, the listening time is one minute. Further, sf used for calculating the listening experience value FE is specified by a pointer (see FIG. 4) associated with the song ID.

Then, the frequency characteristic estimation unit 31 refers to the listening experience value / frequency characteristic correlation table shown in FIG. 5 and determines an individual frequency characteristic (also referred to as “filter type”) FP from the calculated experience value FE. The listening experience value / frequency characteristic correlation table shown in FIG. 5 uses, for example, a statistically calculated correlation between an auditory test for obtaining frequency characteristics of a large number of users and the user's music listening experience. In FIG. 5, the listening experience value FE indicates that the listening experience is small when FE <n1, and the listening experience is large when FE> n2. In addition, when n1 ≦ FE ≦ n2, it indicates that there is a moderate listening experience. Further, as shown in FIG. 5, the frequency characteristic (also referred to as filter type) is a filter having a wider band as the listening experience increases. In other words, this indicates that the more listening experience is, the more the frequency characteristics can be heard over a wide band. This individual frequency characteristic FP constitutes a part of the filter value information input to the feature quantity extraction unit 12.
<1-5> Rhythm characteristic estimation unit The rhythm characteristic estimation unit 32 uses the listening history of the management table managed by the resolution / history management unit 33 and uses the relational expression [Equation 2] to listen to the rhythm characteristic. Listening experience value RE, which is a value indicating the sum of the above, is calculated.

ここにおいて、ｉは楽曲のＩＤ（曲ＩＤ）、ｓｒはその楽曲のリズム成分に関連する特徴量、ｔはその楽曲の聴取時間、ａはその楽曲を聴取した際の注意情報を示す。また、聴取経験値ＲＥの算出に用いるｓｒは、曲ＩＤに対応付けされたポインタ（図４参照）により特定される。

Here, i is a song ID (song ID), sr is a feature quantity related to the rhythm component of the song, t is the listening time of the song, and a is caution information when listening to the song. Further, sr used for calculating the listening experience value RE is specified by a pointer (see FIG. 4) associated with the song ID.

Then, the rhythm characteristic estimation unit 32 refers to the listening experience value / rhythm characteristic correlation table shown in FIG. 6 and determines the individual rhythm characteristic RP from the calculated experience value RE. The listening experience value / rhythm characteristic correlation table shown in FIG. 6 uses, for example, a statistically calculated correlation between a test for obtaining rhythm characteristics of a large number of users and the user's music listening experience. The test for obtaining the user's rhythm characteristic is, for example, preparing a plurality of music pieces with varying rhythm pattern complexity and causing the user to imitate or to distinguish the differences. In FIG. 6, the listening experience value RE indicates that the listening experience is small when RE <n1, and the listening experience is large when RE> n2. In addition, when n1 ≦ RE ≦ n2, it indicates that there is a moderate listening experience. Also, as shown in FIG. 6, the rhythm characteristic has a higher value as the experience increases. In other words, the more listening experience, the more detailed notes and rhythm patterns can be heard. This individual rhythm characteristic RP constitutes part of the filter value information input to the feature quantity extraction unit 12.
<1-6> Attention Estimation Unit The configuration of the attention estimation unit 22 is shown in FIG.

  In FIG. 7, the determination unit 22 a determines whether the user has obtained a result (brain measurement information) of the state of the user's brain with a brain measurement device (not shown) such as near infrared spectroscopy (NIRS) or electroencephalogram (EEG). It is determined whether or not the music content is being listened to, such as whether the music content is being listened to or whether the music content is focused on vocals or musical instruments. The selection unit 22b presents options related to the current state of caution such as “attention to vocals” and “hearing” to the user with buttons and voices on the screen, and receives selection input by the user. The attention estimation unit 22 outputs the attention information input from the determination unit 22a or the attention information input from the selection management unit 22b. As shown in the resolution / history management table shown in FIG. 4, this caution information is indicated by a predefined integer value. In the example shown in FIG. 4, “A (= 1)” is heard, “B (= 2)” is focused on vocals, and “C (= 3)” is focused on the composition of the entire music. . This caution information is used as the value of the parameter a in the above [Equation 1] and [Equation 2].

  The attention information output from the attention estimation unit 22 is stored in a resolution / history management unit 33 provided in the resolution estimation unit 30. Further, the attention information includes the attention information a used in the calculation of the listening experience value FE related to the frequency in the frequency characteristic estimation unit 31 and the calculation of the listening experience value RE related to the rhythm in the rhythm characteristic estimation unit 32. Used as a parameter to indicate. Then, the listening experience value is increased when the user is listening to the content with caution, and conversely, when the user is not paying much attention, the listening experience value is added to be small.

Finally, in order to show the overall relationship between the components described above, FIG. 8 shows a configuration diagram of the change correspondence type preference estimation device in the embodiment for music contents.
The feature amount extraction unit 12, the preference estimation unit 11, the resolution estimation unit 30, and the attention estimation unit 22 in FIG. 8 correspond to the configurations represented by the same names in FIG.
<2> Operation <2-1> Overall Operation The operation of the change correspondence type preference estimation apparatus according to the present embodiment will be described with reference to the flowchart shown in FIG.

First, content to be viewed is read into the apparatus from an external playback / display mechanism (not shown) such as a music playback player or television monitor (step S1).
Subsequently, the feature amount extraction unit 12 extracts a feature amount by applying an individual resolution filter for each user (step S2). The process of extracting feature amounts will be described in detail in <2-2>.

  Thereafter, the preference determination unit 11 determines the user's preference (that is, whether the user likes the content) from the feature amount extracted by the feature amount extraction unit 12, and outputs it as a preference determination result (step S3).

  Next, the resolution estimation unit 30 calculates a value (also referred to as filter value information or resolution filter identification number) indicating the resolution corresponding to the user based on the feature amount extracted by the feature amount extraction unit 12. (Step S4). A method of calculating a value indicating the resolution (filter value information) will be described in detail later.

  Then, after calculating the value indicating the resolution, the attention estimation unit 22 displays the attention information indicating what the content is being watched by, for example, the base melody line or vocal voice quality if the content is viewed. It is estimated (step S5).

  Finally, in the management table in the resolution estimation unit 30, as a new history, the feature amount extracted by the feature amount extraction unit 12 and the value indicating the resolution at that time calculated by the resolution estimation unit 30 (resolution filter identification) Number), the attention information estimated by the attention estimation unit 22 and the listening time are added (step S6), and the process ends. The contents of the attention information and the contents of the resolution / history management table used by the resolution estimation unit 30 are as shown in FIG.

  As shown in FIG. 4, when new content is input, the value of resolution information applied by the resolution application unit 12a is calculated every time the content is viewed, and the management table stores the calculated resolution information. It is updated with the value. Accordingly, a more accurate preference determination result can be output based on the latest resolution information value for the input content to be viewed.

The details of the configuration of the change correspondence type preference estimation apparatus according to the present embodiment will be described below. Here, a case of music content will be described as an example.
<2-2> Operation of Feature Quantity Extraction Unit The operation of the feature quantity extraction unit 12 will be described with reference to the flowchart of FIG.

  First, when a music signal of music content reproduced by a music reproducing device (not shown) is input to the feature quantity extraction unit 12, a frequency spectrum is calculated by the frequency spectrum calculation unit 12b (step S121). The frequency spectrum calculation is performed by a general method such as Fourier transform, for example.

  Next, the latest value of the individual frequency characteristic FP and the value of the individual rhythm characteristic RP managed by the resolution / history management unit 33 are read, and the individual frequency characteristic and the individual rhythm characteristic are obtained as a resolution filter based on these values. Apply (step S122).

Subsequently, the feature amount extraction unit 12 performs a calculation using the individual frequency characteristic and the individual rhythm characteristic on the music signal input using the set value, thereby feeling the user's current resolution. A music feature amount is calculated (step S123). Typical music feature values include tonality as shown in the above-mentioned Patent Document 2, and rhythm patterns as shown in Non-Patent Document 1 and Non-Patent Document 2. The feature amount calculation processing when the rhythm pattern is taken as an example will be described in detail in <2-3>.
<2-3> Rhythm pattern calculation processing FIG. 11 shows a flowchart of the rhythm pattern calculation processing.

  First, based on the individual rhythm characteristics RP for each user, the feature quantity extraction unit 12 determines how much the rhythm pattern of the note value is analyzed in the beat analysis performed when calculating the rhythm pattern. In order to specify, the number of subdivisions is set (step S124). Users with individual rhythm characteristics RP exceeding a certain value can understand complex beat patterns, so that the number of subdivisions is large, that is, subdivision that can analyze fine beat structures such as 32 beats. Set to number of times.

Subsequently, the feature amount calculation unit 12 calculates a beat period and intensity, which are feature amounts representing the beat structure, in response to the music signal input (steps S125 to S127).
Next, the feature quantity extraction unit 12 determines whether or not the number of subdivisions for each user set in step S124 has been reached (step S128). If the number of subdivisions has been reached, the process ends (step S128 is changed). N) If the number of subdivisions has not been reached (step S128 is Y), the processing of steps S125 to S127 is repeated again. As a method of determining the number of subdivisions performed by the feature quantity extraction unit 12, for example, the length of a note value of one beat of a 32nd note when the music playback speed is 60 BPM is set to 125 msec. There is a method in which the preset number of subdivisions is reached when the length of the sound value of 1 beat becomes 125 msec or less.
<2-4> Frequency Characteristic Estimation Process The frequency characteristic estimation process performed by the frequency characteristic estimation unit 31 will be described with reference to the flowchart shown in FIG.

  First, using the information of the listening history of the resolution / history management table managed by the resolution / history management unit 33, a value indicating the sum total of listening experiences regarding the frequency using the above-described relational expression [1]. A certain listening experience value FE is calculated (step S311).

Next, the frequency characteristic estimator 31 refers to the listening experience value / frequency characteristic correlation table, and determines the value of the individual frequency characteristic FP (the identification number of the resolution filter related to the frequency characteristic) from the calculated experience value FE (step). S312).
<2-5> Rhythm characteristic estimation process The rhythm characteristic estimation process performed by the rhythm characteristic estimation unit 32 will be described with reference to the flowchart shown in FIG.

  First, using the information of the listening history of the resolution / history management table managed by the resolution / history management unit 33, the value indicating the total listening experience related to the rhythm characteristics using the relational expression of [Formula 2] described above. A certain listening experience value RE is calculated (step S321).

Next, the rhythm characteristic estimation unit 32 refers to the listening experience value / rhythm characteristic correlation table, and determines the value of the individual rhythm characteristic RP (the identification number of the resolution filter relating to the rhythm characteristic) from the calculated experience value RE (step) S322).
<2-6> Management Table Update Processing The resolution / history management table update processing (management table update processing) managed by the resolution / history management unit 33 will be described with reference to the flowchart shown in FIG.

  First, the value of the individual frequency characteristic FP in the resolution / history management table managed by the resolution / history management unit 33 is updated to the latest value calculated by the frequency characteristic estimation unit 31 (step S331).

  Next, the value of the individual rhythm characteristic RP in the resolution / history management table managed by the resolution / history management unit 33 is updated to the latest value calculated by the rhythm characteristic estimation unit 32 (step S332).

Subsequently, the music ID, the music feature amount calculated by the feature amount extraction unit 12, the individual frequency characteristic FP at the time of listening, the individual rhythm characteristic RP, and the attention information at the time of listening estimated by the attention estimation unit 22 are resolution / history management. It is added to the history information of the table (step S333).
<Modification>
(1) In the above-described embodiment, the content has been described as music. However, the content may include video information such as photographs, paintings, and videos, and sound information other than music. When the content is a photograph, the feature amount representing the feature of the content used in FIG. 8 may be changed to a SIFT feature amount, a color feature amount, or the like.

  (2) In the above-described embodiment, the resolution application unit 12a described the calculation based on the two types of the individual frequency characteristic and the individual rhythm characteristic as means for determining the individual characteristic. However, the present invention is not limited to this. For example, the characteristics may be determined using other means such as individual timbre characteristics indicating the characteristics of individual musical instruments with respect to the timbre.

  (3) In the above-described embodiment, when the resolution application unit 12a calculates the individual frequency characteristics and the individual rhythm characteristics of each user, the listening experience value / rhythm characteristic correlation table shown in FIG. Although an example using the listening experience value / rhythm characteristic correlation table created based on the experience is described, it is not limited to this. For example, it may be calculated by performing a response characteristic analysis on sound by brain measurement.

  (4) In the above-described embodiment, the example in which the frequency characteristic and the rhythm characteristic are used as the characteristic amount for analyzing music has been described, but the present invention is not limited to this. For example, tonality, chord appearance frequency, volume-related feature quantities (for example, volume distribution, average volume, maximum volume, etc.), music composition (for example, bar structure, etc.), timbre composition (for example, instrument structure, etc.) are used. There may be.

  (5) Recording a program consisting of a processor of the change correspondence type preference estimation apparatus shown in the above-described embodiment and a program code to be executed by various circuits connected to the processor on a recording medium or various communication paths It is possible to distribute and distribute it through the like. Examples of such a recording medium include an IC card, a hard disk, an optical disk, a flexible disk, and a ROM. The distributed and distributed control program is used by being stored in a memory or the like that can be read by the processor, and the processor executes the control program to realize the functions shown in the embodiments. Become so. A part of the control program is transmitted to a device (processor) capable of executing a program separate from the image management apparatus via various networks, and the part of the control program is executed in the separate program-executable device. It is also possible to make it.

  (6) A part or all of the constituent elements of the change correspondence preference estimation device described in the above embodiment may be mounted as one or a plurality of integrated circuits (IC, LSI, etc.) Other elements may be added to the components of the image management apparatus to form an integrated circuit (one chip).

  The name used here is LSI, but it may also be called IC, system LSI, super LSI, or ultra LSI depending on the degree of integration. Further, the method of circuit integration is not limited to LSI's, and implementation using dedicated circuitry or general purpose processors is also possible. An FPGA (Field Programmable Gate Array) that can be programmed after manufacturing the LSI, or a reconfigurable processor that can reconfigure the connection or setting of circuit cells inside the LSI may be used. Further, if integrated circuit technology comes out to replace LSI's as a result of the advancement of semiconductor technology or a derivative other technology, it is naturally also possible to carry out function block integration using this technology. Biotechnology can be applied.

  The change correspondence type preference estimation apparatus and the change correspondence type preference estimation method according to the present invention have a resolution application unit, a resolution estimation unit, and a caution estimation unit, and are useful for content search, automatic content configuration, content recommendation, and the like.

DESCRIPTION OF SYMBOLS 11 Preference estimation part 12 Feature quantity extraction part 12a Resolution application part 12b Frequency spectrum calculation part 12c Feature quantity extraction part 22 Attention estimation part 22a Judgment part 22b Selection management part 30 Resolution estimation part 31 Frequency characteristic estimation part 31 Rhythm characteristic estimation part 33 Resolution・ History management department

Claims (7)

A feature amount extraction unit for extracting features of content;
A resolution application unit that applies a filter corresponding to the resolution for each user in the feature amount extraction unit;
A change correspondence type preference estimation device comprising: a resolution estimation unit for changing the filter value of the resolution application unit in accordance with a user's content appreciation experience. The change correspondence type preference estimation device further includes:
The change correspondence type preference estimation apparatus according to claim 1, further comprising a attention estimation unit for changing a change characteristic of the filter value of the resolution estimation unit depending on how the user pays attention to the content. . The said resolution estimation part has all the frequency characteristic estimation part which estimates the frequency resolution of a user's auditory characteristic, the rhythm characteristic estimation part which estimates rhythm resolution, or any one or more. The change correspondence type preference estimation apparatus according to claim 2. The change correspondence type preference estimation device further includes:
The resolution estimation unit includes a listening history management unit that manages a listening history including a feature amount of content after filtering by the resolution application unit that the user has viewed in the past, and attention information at the time of listening estimated by the attention estimation unit. The change correspondence type preference estimation apparatus according to any one of claims 1 to 3, further comprising: A change correspondence type preference estimation method in a change correspondence type preference estimation device for estimating a user's preference for content,
A feature extraction step for extracting features of the content;
A resolution application step of applying a filter corresponding to the resolution for each user in the feature amount extraction unit;
A resolution estimation step for changing the filter value of the resolution application unit in accordance with a user's content appreciation experience. A change correspondence type preference estimation program for causing a computer to function as a change correspondence type preference estimation device for estimating a user's preference for content,
In the computer,
A feature extraction step for extracting features of the content;
A resolution application step of applying a filter corresponding to the resolution for each user in the feature amount extraction unit;
And a resolution estimation step for changing the filter value of the resolution application unit according to the user's content appreciation experience. A feature amount extraction unit for extracting features of content;
A resolution application unit that applies a filter corresponding to the resolution for each user in the feature amount extraction unit;
A change correspondence type preference estimation integrated circuit comprising: a resolution estimation unit configured to change a filter value of the resolution application unit according to a user's content appreciation experience.

Images