Logo image
Back
Gender detection based on spatial pyramid matching
Conference proceeding   Peer reviewed

Gender detection based on spatial pyramid matching

Fahimeh Alaei and Alireza Alaei
Document Analysis and Recognition – ICDAR 2021, pp.305-317
Lecture Notes in Computer Science
International Conference on Document Analysis and Recognition (Lausanne, Switzerland, 05/12/2021–10/12/2021)
02/09/2021
DOI: https://doi.org/10.1007/978-3-030-86337-1_21
url
Gender detection based on spatial pyramid matchingView
Published (Version of record)Conference paper

Related links

Metrics

79 Record Views
4 Times Cited - Web of Science

Abstract

Gender detection Handwritten document images Spatial pyramid matching
The similarity and homogeneous visual appearance of male and female handwriting make gender detection from off-line handwritten document images a challenging research problem. In this paper, an effective method based on spatial pyramid matching is proposed for gender detection from handwritten document images. In the proposed method, the input handwritten document image is progressively divided into several sub-regions from coarse to fine levels. The weighted histograms of the sub-regions are then calculated. This process is resulting in a spatial pyramid feature set which is an extension of the orderless bag-of-features image representation. Classical classifiers, such as Support Vector Machines and ensemble classifiers, are considered for determining the gender (male and female) of individuals from their handwriting. Experiments were conducted on two benchmarks, QUWI and MSHD datasets, and the proposed method provided a promising improvement in gender detection accuracies, especially in script-dependent scenarios, compared with the results reported in the literature.

Details

Logo image