Forensic interpretation framework for body and gait analysis: feature extraction, frequency and distinctiveness

ABSTRACT

Surveillance is ubiquitous in modern society, allowing continuous monitoring of areas that results in capturing criminal (or suspicious) activity as footage. This type of trace is usually examined, assessed and evaluated by a forensic examiner to ultimately help the court make inferences about who was on the footage. The purpose of this study was to develop an analytical model that ensures applicability of morphometric (both anthropometric and morphological) techniques for photo-comparative analyses of body and gait of individuals in CCTV images, and then to assign a likelihood ratio. This is the first paper of a series: This paper will contain feature extraction to observe repeatability procedures from a single observer, in turn, producing the frequency and distinctiveness of the feature set within the given population. To achieve this, an Australian population database of 383 subjects (stance) and 268 subjects (gait) from both sexes, all ages above 18 and ancestries was generated. Features were extracted, defined, and their rarity viewed among the developed database. Repeatability studies were completed in which stance and gait (static and dynamic) features contained low levels of repeatability error (0.2%–1.5 TEM%). For morphological examination, finger flexion and feet placement were observed to have high observer performance.

KEYWORDS:

• Forensic gait analysis
• forensic interpretation
• forensic morphometric examination

View correction statement:

Correction

Acknowledgments

The authors would like to acknowledge all the subjects who volunteered in this study and were photographed/filmed under ethics approval at the University of Technology Sydney. Without their generous contribution, this study would not be possible.

Ethics

Under ethics approval (UTS HREC Ref No. 201500051), subjects were recruited and photographed/filmed. All subject data were anonymized, and all personal information was stored separate to number coded images (in a PIN access room, on a password-protected computer and in a locked safe). All CCTV footage obtained was also stored under the same conditions.

Correction Statement

This article was originally published with errors, which have now been corrected in the online version. Please see Correction http://doi.org/10.1080/00450618.2024.2377287

Notes

^a. Ground truth by Cardoso et al.,³² I is defined as ‘the reference values used as standard for comparison purposes’.

^b. Validity is defined by Meuwly et al.,³³ as ‘range of conditions for which the method has been tested’.

^c. An analytical model in the context of this paper is the use of a defined set of morphometric measurements that are robust to the forensic environment.

^d. A plumb line is a cord with a weighted plumb attached to provide a vertical line, dividing the body into two (coronal and/or sagittal).²⁰ A virtual plumb line was also added, which applied the same concepts to that with the weighted plumb, dividing the body into two.

^e. By dividing the anthropometric measurement by the total sum of all measurements, indices are attained, subsequently proportions will be compared instead of sizes: Indices = $\frac{\mathit{Anthropometric}\mathit{Measurement}}{\mathit{Total}\sum \mathit{of}\mathit{all}\mathit{the}\mathit{Measurements}}\mathit{\hspace{0.17em}}$.