Eureka AIR delivers breakthrough ideas for toughest innovation challenges, trusted by R&D personnel around the world.

A Cross-media Retrieval Method Based on Markov-like Correlation Measure

A correlation measurement and cross-media technology, applied in the field of cross-media retrieval, can solve problems such as slow retrieval speed, large amount of cross-media retrieval data, and high algorithm complexity, and achieve the effect of reducing time complexity

Active Publication Date: 2020-07-14
XIANGTAN UNIV
View PDF2 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0004] Since cross-media retrieval involves the extraction of the underlying feature data of different media and the data association between different modalities, there are the following technical problems: 1) The selection of feature extraction dimensions for different data sets is a difficult problem; 2) Different modalities The dimensions after media feature extraction are different, and the "dimension disaster" can be solved by unifying into the same subspace; 3) Due to the large amount of cross-media retrieval data, the traditional European distance correlation measurement method has high algorithm complexity and slow retrieval speed

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • A Cross-media Retrieval Method Based on Markov-like Correlation Measure

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0032]In order to illustrate the present invention more clearly, the Wikipedia dataset is used as the data source to search, and the ideas of the present invention are elaborated more deeply and specifically; the Wikipedia dataset contains two media, text and images, and the dataset is composed of , history, geography and art and other 10 categories, the text is some news reports about the content of these categories, and the image is a picture related to the content. The Wikipedia dataset contains a total of 2866 samples.

[0033] In conjunction with accompanying drawing, the concrete steps of the present invention are as follows:

[0034] Step 1: Input the Wikipedia dataset database, which contains 2866 pairs of text and images.

[0035] Step 2: Store the text and images in the text and image databases respectively in the Wikipedia dataset image dataset folder and the Wikipedia dataset text dataset folder under the Wikipedia dataset, and store the Wikipedia dataset image dat...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention discloses a cross-media retrieval method based on the martensite-like relativity measurement. The method comprises the following steps that a text database and an image database are input; texts and the images are both divided into a training set and a testing set; feature extraction is separately conducted on the training set and the testing set of the texts and on the training set and the testing set of the images, and a training feature set and a testing feature set of the texts and a training feature set and a testing feature set of the images are obtained respectively; data of the feature sets of the texts and the images is unified into the same subspace; after the data of the feature sets of the texts and the images is unified into the same subspace, the relativity between the testing set of the texts and the testing set of the images is calculated by using the martensite-like distance formula; according to the classifications and the mahalanobis distances which the texts and the images belong to respectively, the retrieval precision is determined by using an average precision index and a recall rate index. The maximal common feature subspace of the feature data sets of the texts and the images is found through the W-CCA algorithm, the training set and the testing set of the texts and the training set and the testing set of the images are mapped to the feature subspace, so that the data of the feature sets of the texts and the images is unified into the same subspace, the problem about curse of dimensionality is solved, and the most prominent features of original data are maximally retained.

Description

technical field [0001] The invention relates to a cross-media retrieval method, in particular to a cross-media retrieval method based on Markov-like correlation measure. Background technique [0002] Cross-media retrieval is a new concept proposed in recent years. Traditional retrieval methods are basically single-modal retrieval, that is, text retrieval for text, image retrieval for images, audio retrieval for audio or video retrieval for video. The most typical representatives are Baidu, Google , yahoo and other commercial public information retrieval engines and music and video retrieval platforms such as cool dog and youtube are all implemented on the basis of text annotation. Because of market demand, cross-media retrieval has become a research hotspot in recent years. On the basis of multimedia, cross-media uses the forms and characteristics of various media to process the same or related information with different media expressions, so as to achieve the purposes of s...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Patents(China)
IPC IPC(8): G06F16/41G06F16/432G06F16/48
CPCG06F16/41G06F16/434G06F16/48
Inventor 裴廷睿吴海滨赵津锋曹江莲田淑娟
Owner XIANGTAN UNIV
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Eureka Blog
Learn More
PatSnap group products