A method and device for obtaining correlation between multimedia data

Abstract

The invention provides a method and device obtaining multimedia data correlation, relates to the computer technical field, and solves the problems that an existing terminal cannot consider user operations and specific application scenes when calculating audio frequency correlations, and thus providing incomplete audio frequency content correlation data for an intelligent recommend system; the novel method comprises the following steps: gathering each user operation digit identification so as to obtain a matrix R according to multimedia data attribute information and preset user operations; clustering the matrix R so as to obtain all matrixes R', wherein the matrixes R' are the sub matrixes of the matrix R; decomposing each matrix R' in sequence so as to obtain a matrix S for each matrix R', wherein a matrix S element Svu represents a weight value of the multimedia data v belonging to the multimedia data type u; calculating a first correlation degree according to all matrixes S comprising first multimedia data identification and second multimedia data identification.

Description

technical field

[0001] The invention relates to the field of computer technology, in particular to a method and device for acquiring the correlation between multimedia data. Background technique

[0002] Nowadays, in this era of rapid Internet development, people's demand for audio-visual aspects is getting higher and higher, and the audio-related recommendation service can recommend other tracks that the user may like for the user according to the user's personalized preference, effectively helping the user discover the demand , to promote users' demand for audio-visual services.

[0003] At present, when a terminal associates audios, it usually associates audios in a content-based manner: that is, the terminal directly calculates the audio correlation between the audios according to the inherent tags of each audio. However, this association method only associates the audio content of the audio, and does not consider the user's operation behavior and specific application s...

Examples

Embodiment 1

[0045] Exemplarily, when the accumulated value of all elements in the column of each multimedia type of matrix R in step 102 decreases from left to right and the accumulated value of all elements in the row of each user identifier in the matrix R is given by When descending in order from top to bottom, the device for obtaining the correlation between multimedia data in step 103 clusters the matrix R, and the specific clustering process of obtaining a matrix R' can be realized through the following two implementation methods:

[0046] The first implementation mode, according to the expansion mode of expanding the lower right corner of the matrix R first row and then column, any clustering process in step 103 includes the following steps:

[0047] 103a1. The device for obtaining the correlation between multimedia data obtains the sub-matrix R1 from the matrix R, and divides the sum of all elements in the matrix R1 by the number of elements in the matrix R1 to obtain the user oper...

Embodiment 2

[0067] Exemplarily, when the accumulated value of all elements in the column of each multimedia type of matrix R in step 102 increases sequentially from left to right and the accumulated value of all elements in the row of each user identifier in the matrix R is given by When increasing sequentially from top to bottom, the device for obtaining the correlation between multimedia data in step 103 clusters the matrix R, and the specific clustering process of obtaining a matrix R' can be realized in the following two ways:

[0068] The first implementation mode, according to the expansion mode of expanding the upper right corner of the matrix R first row and then column, any clustering process in step 103 includes the following steps:

[0069]103c1. The device for obtaining the correlation between multimedia data obtains the sub-matrix R1' from the matrix R, and divides the sum of all elements in the matrix R1' by the number of elements in the matrix R1' to obtain the user's respon...

Embodiment 3

[0079] Exemplarily, the matrix R shown in Table 2 is taken here as an example. When clustering the matrix R, if the clustering is performed according to the expansion method of expanding to the lower right corner of the matrix R first row and then column, then the clustering The process is as follows:

[0080] First, select any element from the matrix R as the initial element of the first clustering. For example, if U2B1 in the first row and first column is used as the initial point, the first clustering process of the matrix R As follows:

[0081] First, starting from U2B1, calculate the operation density value of a single U2 to B1 based on the user operation density formula, and the density value is 2 / (1)=2;

[0082] Second, since the clustering process is a process of clustering media data types belonging to the same category, the number of users and the number of multimedia data types must be greater than 1 when clustering starts, so add the second column directly at this...

technical field

[0001] The invention relates to the field of computer technology, in particular to a method and device for acquiring the correlation between multimedia data. Background technique

[0002] Nowadays, in this era of rapid Internet development, people's demand for audio-visual aspects is getting higher and higher, and the audio-related recommendation service can recommend other tracks that the user may like for the user according to the user's personalized preference, effectively helping the user discover the demand , to promote users' demand for audio-visual services.

[0003] At present, when a terminal associates audios, it usually associates audios in a content-based manner: that is, the terminal directly calculates the audio correlation between the audios according to the inherent tags of each audio. However, this association method only associates the audio content of the audio, and does not consider the user's operation behavior and specific application s...

Examples

Embodiment 1

[0045] Exemplarily, when the accumulated value of all elements in the column of each multimedia type of matrix R in step 102 decreases from left to right and the accumulated value of all elements in the row of each user identifier in the matrix R is given by When descending in order from top to bottom, the device for obtaining the correlation between multimedia data in step 103 clusters the matrix R, and the specific clustering process of obtaining a matrix R' can be realized through the following two implementation methods:

[0046] The first implementation mode, according to the expansion mode of expanding the lower right corner of the matrix R first row and then column, any clustering process in step 103 includes the following steps:

[0047] 103a1. The device for obtaining the correlation between multimedia data obtains the sub-matrix R1 from the matrix R, and divides the sum of all elements in the matrix R1 by the number of elements in the matrix R1 to obtain the user oper...

Embodiment 2

[0067] Exemplarily, when the accumulated value of all elements in the column of each multimedia type of matrix R in step 102 increases sequentially from left to right and the accumulated value of all elements in the row of each user identifier in the matrix R is given by When increasing sequentially from top to bottom, the device for obtaining the correlation between multimedia data in step 103 clusters the matrix R, and the specific clustering process of obtaining a matrix R' can be realized in the following two ways:

[0068] The first implementation mode, according to the expansion mode of expanding the upper right corner of the matrix R first row and then column, any clustering process in step 103 includes the following steps:

[0069]103c1. The device for obtaining the correlation between multimedia data obtains the sub-matrix R1' from the matrix R, and divides the sum of all elements in the matrix R1' by the number of elements in the matrix R1' to obtain the user's respon...

Embodiment 3

[0079] Exemplarily, the matrix R shown in Table 2 is taken here as an example. When clustering the matrix R, if the clustering is performed according to the expansion method of expanding to the lower right corner of the matrix R first row and then column, then the clustering The process is as follows:

[0080] First, select any element from the matrix R as the initial element of the first clustering. For example, if U2B1 in the first row and first column is used as the initial point, the first clustering process of the matrix R As follows:

[0081] First, starting from U2B1, calculate the operation density value of a single U2 to B1 based on the user operation density formula, and the density value is 2 / (1)=2;

[0082] Second, since the clustering process is a process of clustering media data types belonging to the same category, the number of users and the number of multimedia data types must be greater than 1 when clustering starts, so add the second column directly at this...

Images