Authors
Mullai Malar K, Mrs. Kajal Vinayakrao Waghmare, Krushna Sharad Sonawane
Abstract
In situations where traditional identification techniques, such as facial recognition, frequently fall short, this study investigates the use of gait pattern analysis as a reliable forensic technique for identifying individuals from low-resolution or blurry CCTV footage. To improve footage quality and extract useful gait data, the study incorporates artificial intelligence (AI)-based video augmentation using technologies such as Pixelcut AI and deep learning models. Gait characteristics, including joint angles and motion trajectories, were retrieved and examined using pose estimation tools such as Kinovea. For precise individual identification, a CNN-LSTM hybrid model was created to categorize and match gait patterns. Real-world grocery surveillance film and public gait datasets were used to validate the methodology, guaranteeing data anonymization and ethical compliance. The results confirmed the feasibility of AI-enhanced gait analysis in forensic situations, with a 91.3% accuracy in identifying persons. Concerns about legal admissibility, dataset constraints, and computational complexity are among the difficulties that are highlighted in the study. Even in situations where video is obscured or damaged, the study demonstrates that gait analysis is a reliable, consistent, and non-intrusive biometric method. It promises future opportunities for multi-biometric systems, real-time surveillance integration, and the creation of extensive, varied gait datasets. It also provides a potent substitute for facial recognition in forensic investigations.
Closed-circuit television (CCTV) systems have proliferated and are now a vital component of contemporary security infrastructure, supporting forensic investigations as well as crime prevention. However, poor video quality, often caused by issues such as low resolution, inadequate lighting, and motion blur, frequently undermines the effectiveness of these devices. Traditional identification techniques, such as facial recognition, which primarily rely on unambiguous visual inputs, are severely hindered by these issues. On the other hand, gait analysis, which examines each person's distinct walking style, provides a robust substitute. A useful tool in forensic research, gait is regarded as a unique and essentially unchangeable biometric characteristic that may be examined even in video situations that are degraded [1].
In the past, gait analysis was limited to clinical settings, when walking patterns were evaluated using instrumented gait mats and marker-based motion capture devices. Despite their accuracy, these approaches required a lot of resources and were impractical for general use. This discipline has undergone a revolution with the introduction of computer vision and artificial intelligence (AI), which have made it possible to extract gait parameters from basic video recordings. The potential to do quantitative gait analyses using inexpensive equipment, like cellphones, has been shown in recent studies, expanding the use of gait analysis in a variety of fields, including forensics [2].
A number of strong arguments support the choice of gait analysis for person recognition in hazy CCTV footage,
i. Resilience in Low-Quality Video: It is frequently possible to extract gait patterns from blurry or low-resolution movies where it is difficult to distinguish facial features. Gait analysis is especially useful in forensic situations with less-than-ideal film because of its durability.
ii. Non-Intrusive and Covert: Gait analysis may be carried out covertly, which makes it appropriate for forensic and surveillance applications in contrast to other biometric techniques that necessitate subject involvement.
iii. Complementary to Other Biometrics: Combining gait analysis with other biometric techniques, such facial recognition, can improve identification precision overall and give forensic investigators a more complete toolkit [1].
References
Patel, R., Sharma, V., & Mehta, S. (2022). Enhancing blurry CCTV footage for forensic analysis using AI-based video restoration. Forensic Science International, 335, 111237. https://doi.org/10.1016/j.forsciint.2022.111237
Singh, A., Kumar, R., & Gupta, P. (2017). A review on biometric gait recognition: Approaches, datasets, and challenges. Neurocomputing, 275, 270-284. https://doi.org/10.1016/j.neucom.2017.05.041
Kharb, A., Saini, V., Jain, Y. K., & Dhiman, S. (2011). A review of gait cycle and its parameters. IJCEM International Journal of Computational Engineering & Management, 13, 78–83. Retrieved from http://www.ijcem.org
Birch, I., Raymond, L., Christou, A., & Fernando, M. A. (2013). The identification of individuals by observational gait analysis using closed-circuit television footage. Science & Justice, 53(3), 339–342. https://doi.org/10.1016/j.scijus.2013.04.005
T. Hwang and A. Effenberg, "Gait Analysis: Head Vertical Movement Leads to Lower Limb Joint Angle Movements," 2022 IEEE International Conference on Consumer Electronics (ICCE), Las Vegas, NV, USA, 2022, pp. 1-5, doi:10.1109/ICCE53296.2022.9730350.
Bao, T., Gao, J., Wang, J., Chen, Y., Xu, F., Qiao, G., & Li, F. (2023). A global bibliometric and visualized analysis of gait analysis and artificial intelligence research from 1992 to 2022. Frontiers in Robotics and AI, 10. https://doi.org/10.3389/frobt.2023.1265543
Gaud, N., Rathore, M., & Suman, U. (2023). Human gait analysis and activity recognition: A review. 2023 IEEE Guwahati Subsection Conference (GCON), 23–25 June 2023, Guwahati, India. IEEE. https://doi.org/10.1109/GCON58516.2023.10183458
Gaud, N., Rathore, M., & Suman, U. (2023). Human gait analysis and activity recognition: A review. 2023 IEEE Guwahati Subsection Conference (GCON). DOI:10.1109/GCON58516.2023.10183458
Kumar, S., et al. "A Review of Deep Learning Approaches for Human Gait Recognition." ResearchGate, July 2023. https://doi.org/10.1109/AISC56616.2023.10085665
Mangone, M., Marinelli, E., Santilli, G., Finanore, N., Agostini, F., Santilli, V., Bernetti, A., Paoloni, M., & Zaami, S. (2023). Gait analysis advancements: rehabilitation value and new perspectives from forensic application. European Review for Medical and Pharmacological Sciences, 27(1), 3–12. https://doi.org/10.26355/eurrev_202301_30847
Aung, S. T. Y., &Kusakunniran, W. (2024). A comprehensive review of gait analysis using deep learning approaches in criminal investigation. PeerJ Computer Science, 10, e2456. https://doi.org/10.7717/peerj-cs.2456
Özateş, M. E., Yaman, A., Salami, F., Campos, S., Wolf, S. I., & Schneider, U. (2024). Identification and interpretation of gait analysis features and foot conditions by explainable AI. Scientific Reports, 14, Article 5998. https://doi.org/10.1038/s41598-024-42064-2
Chatterjee, S., Kumar, A., & Roy, S. (2024). Emerging trends in gait recognition based on deep learning: a survey. PeerJ Computer Science, 10, e2158. https://doi.org/10.7717/peerj-cs.2158
R. Kumar et al., "Detection Of Criminal Activities/Criminal Through CCTV By Live Footage Analysis," 2024 OPJU International Technology Conference (OTCON) on Smart Computing for Innovation and Advancement in Industry 4.0, Raigarh, India, 2024, pp. 1-6, doi:10.1109/OTCON60325.2024.10687573.
Seckiner, D., Mallett, X., Roux, C., Gittelson, S., Maynard, P., &Meuwly, D. (2024). Forensic interpretation framework for body and gait analysis: Feature extraction, frequency, and distinctiveness. Australian Journal of Forensic Sciences, 56(4), 338–354. https://doi.org/10.1080/00450618.2022.2161636
How to cite this article?
| APA Style | Parmar, D., Chavda, B., & Jadav, G. (2026). A comprehensive review of deepfake audio detection: Techniques, applications, and countermeasures. Academic Journal of Forensic Sciences, 9(1), 1–16. |
| Chicago Style | |
| MLA Style | |
| DOI | |
| URL |