Abstract
Gait, the pattern or style in which locomotion is undertaken, has kinematic characteristics that may occur in varying proportions of a population and therefore have discriminatory potential. Forensic gait analysis is the analysis, comparison and evaluation of features of gait to assist the investigation of crime. While there have been recent developments in automated gait recognition systems, gait analysis presented in criminal court to assist in identification currently relies on observational analysis by expert witnesses. Observational gait analysis has been the focus of considerable research, and it has been shown that the adoption of a systematic approach to both the observation and recording of features of gait improves the reliability of the analysis. The Sheffield Features of Gait Tool was developed by forensic gait analysis practitioners based on their casework and trial experience, and consists of more than a hundred features of gait and variances. This paper reports the findings of a study undertaken to assess the repeatability and reproducibility of the Sheffield Features of Gait Tool.
Fourteen participants, with experience in observational gait analysis, viewed footage of computer generated avatars walking, and completed the features of gait tool on multiple occasions. The repeatability scores varied between participants from a highest score of 42.59 out of a maximum possible score of 45 (94.65%), to a lowest score of 30.76 (68.35%), with a mean score of 35.79 (79.54%) and a standard deviation of 3.59 (7.98%). The reproducibility scores for the assessment of each avatar varied from a highest score of 137.73 out of the best possible score of 180 (76.52%), to a lowest score of 127.21 (70.67%), with a mean score of 132.21 (73.45) and a standard deviation of 3.82 (2.12%). The results demonstrated that the use of the Sheffield Features of Gait Tool by experienced analysists resulted in what could be considered to be good levels of both repeatability and reproducibility. Some variation was shown to occur both between the results produced by different analysts, and between those produced from the analysis of different avatars. An understanding of the probative value of gait analysis evidence is an important facet of its submission as evidence, and the design and testing of standardized methods of analysis and comparison are an essential element of developing that understanding. This study is the first to test a purpose designed features of gait tool for use in forensic gait analysis.
Fourteen participants, with experience in observational gait analysis, viewed footage of computer generated avatars walking, and completed the features of gait tool on multiple occasions. The repeatability scores varied between participants from a highest score of 42.59 out of a maximum possible score of 45 (94.65%), to a lowest score of 30.76 (68.35%), with a mean score of 35.79 (79.54%) and a standard deviation of 3.59 (7.98%). The reproducibility scores for the assessment of each avatar varied from a highest score of 137.73 out of the best possible score of 180 (76.52%), to a lowest score of 127.21 (70.67%), with a mean score of 132.21 (73.45) and a standard deviation of 3.82 (2.12%). The results demonstrated that the use of the Sheffield Features of Gait Tool by experienced analysists resulted in what could be considered to be good levels of both repeatability and reproducibility. Some variation was shown to occur both between the results produced by different analysts, and between those produced from the analysis of different avatars. An understanding of the probative value of gait analysis evidence is an important facet of its submission as evidence, and the design and testing of standardized methods of analysis and comparison are an essential element of developing that understanding. This study is the first to test a purpose designed features of gait tool for use in forensic gait analysis.
Original language | English |
---|---|
Pages (from-to) | 544-551 |
Number of pages | 8 |
Journal | Science and Justice |
Volume | 59 |
Issue number | 5 |
DOIs | |
Publication status | Published - 19 Apr 2019 |