Feature Extraction with Computational Intelligence for Head Pose Estimation

Shane Reid; Sonya Coleman; Dermot Kerr; Philip Vance; Siobhan O'Neill

Feature Extraction with Computational Intelligence for Head Pose Estimation

Shane Reid, Sonya Coleman, Dermot Kerr, Philip Vance, Siobhan O'Neill

Research output: Contribution to conference › Paper › peer-review

104 Downloads (Pure)

Abstract

Measuring social signals has often proved challenging as they are often characterized by subtle movements which are difficult to detect. Head pose is one such social signal used to indicate where an individual’s attention is focused. This paper will discuss the problem of head pose estimation by defining the problem in terms of two fields of view, pan and tilt. A novel approach for head pose estimation is described that uses histogram of oriented gradients with support vector machines. The approach is compared with a template matching approach, among others, using a well-known dataset. The results show that the histogram of oriented gradients approach is the most accurate, able to determine head pan to within one class approximately 79% of the time, and head tilt to within one class approximately 82% of the time.

Original language	English
Number of pages	7
Publication status	Accepted/In press - 1 Sept 2018
Event	SYMPOSIUM SERIES ON COMPUTATIONAL INTELLIGENCE: IEEE Symposium on Computational Intelligence in Feature Analysis, Selection and Learning in Image and Pattern Recognition - Benguluru, India Duration: 18 Nov 2018 → 20 Nov 2018

Conference

Conference	SYMPOSIUM SERIES ON COMPUTATIONAL INTELLIGENCE
Abbreviated title	FASLIP
Country/Territory	India
City	Benguluru
Period	18/11/18 → 20/11/18

Keywords

Head Pose estimation
Template matching
Social signal processing
Histogram of oriented gradients
Support vector machine

Access to Document

SSCI_ReidAuthor Accepted Manuscript, 1.56 MB

Cite this

@conference{c48057c58e94407099b587732b39d74e,

title = "Feature Extraction with Computational Intelligence for Head Pose Estimation",

abstract = "Measuring social signals has often proved challenging as they are often characterized by subtle movements which are difficult to detect. Head pose is one such social signal used to indicate where an individual{\textquoteright}s attention is focused. This paper will discuss the problem of head pose estimation by defining the problem in terms of two fields of view, pan and tilt. A novel approach for head pose estimation is described that uses histogram of oriented gradients with support vector machines. The approach is compared with a template matching approach, among others, using a well-known dataset. The results show that the histogram of oriented gradients approach is the most accurate, able to determine head pan to within one class approximately 79% of the time, and head tilt to within one class approximately 82% of the time. ",

keywords = "Head Pose estimation, Template matching, Social signal processing, Histogram of oriented gradients, Support vector machine",

author = "Shane Reid and Sonya Coleman and Dermot Kerr and Philip Vance and Siobhan O'Neill",

year = "2018",

month = sep,

day = "1",

language = "English",

note = "SYMPOSIUM SERIES ON COMPUTATIONAL INTELLIGENCE : IEEE Symposium on Computational Intelligence in Feature Analysis, Selection and Learning in Image and Pattern Recognition, FASLIP ; Conference date: 18-11-2018 Through 20-11-2018",

}

TY - CONF

T1 - Feature Extraction with Computational Intelligence for Head Pose Estimation

AU - Reid, Shane

AU - Coleman, Sonya

AU - Kerr, Dermot

AU - Vance, Philip

AU - O'Neill, Siobhan

PY - 2018/9/1

Y1 - 2018/9/1

N2 - Measuring social signals has often proved challenging as they are often characterized by subtle movements which are difficult to detect. Head pose is one such social signal used to indicate where an individual’s attention is focused. This paper will discuss the problem of head pose estimation by defining the problem in terms of two fields of view, pan and tilt. A novel approach for head pose estimation is described that uses histogram of oriented gradients with support vector machines. The approach is compared with a template matching approach, among others, using a well-known dataset. The results show that the histogram of oriented gradients approach is the most accurate, able to determine head pan to within one class approximately 79% of the time, and head tilt to within one class approximately 82% of the time.

AB - Measuring social signals has often proved challenging as they are often characterized by subtle movements which are difficult to detect. Head pose is one such social signal used to indicate where an individual’s attention is focused. This paper will discuss the problem of head pose estimation by defining the problem in terms of two fields of view, pan and tilt. A novel approach for head pose estimation is described that uses histogram of oriented gradients with support vector machines. The approach is compared with a template matching approach, among others, using a well-known dataset. The results show that the histogram of oriented gradients approach is the most accurate, able to determine head pan to within one class approximately 79% of the time, and head tilt to within one class approximately 82% of the time.

KW - Head Pose estimation

KW - Template matching

KW - Social signal processing

KW - Histogram of oriented gradients

KW - Support vector machine

M3 - Paper

T2 - SYMPOSIUM SERIES ON COMPUTATIONAL INTELLIGENCE

Y2 - 18 November 2018 through 20 November 2018

ER -

Feature Extraction with Computational Intelligence for Head Pose Estimation

Abstract

Conference

Keywords

Access to Document

Fingerprint

Cite this