Police investigating the alleged murder of Pune businessman Ketan Agarwal at Lohagad Fort have incorporated AI-assisted gait analysis into the investigation after surveillance footage failed to provide a clear view of the suspect's face.
Police are investigating the death of Pune businessman Ketan Agarwal, whose body was recovered from a gorge near Lohagad Fort in Pune district, Maharashtra, on June 22, 2026.
Investigators have registered a murder case and arrested Agarwal's fiancée, Siya Goyal, and co-accused Chetan Chaudhary in connection with the incident. Police allege that the death was planned and are examining CCTV footage, digital evidence, and crime scene reconstruction as part of the investigation.
During the investigation, CCTV cameras reportedly captured a hooded individual following the couple before the incident. Because the person's face was not clearly visible, investigators have turned to AI-assisted forensic gait analysis to compare the walking pattern seen in the surveillance footage with that of the accused.
As reported in NDTV, Officials have stated that the findings from gait analysis will be evaluated alongside other forensic and digital evidence collected during the investigation.
L.N. Rao, a former Delhi Police Special Cell officer, said gait analysis can assist investigators by providing corroborative evidence that complements other forms of forensic evidence.
It involves analyzing a suspect's walking pattern, including gait, body posture and the angles of movement. Forensic experts examine these movement patterns to establish crucial facts. While gait analysis is not conclusive evidence on its own, it can significantly strengthen the overall case when supported by other evidence.
L.N. Rao, Former Delhi Police Special Cell Officer
As part of the reconstruction, the accused is expected to wear clothing matching that seen in the CCTV footage and walk along the same route at Lohagad Fort. Investigators will then compare the recreated recording with the original surveillance video to assess similarities in the walking pattern, as reported in NDTV.
The analysis will focus on measurable movement characteristics, including stride length, walking speed, posture, arm swing, limb coordination, and overall body mechanics.
Police also plan to use artificial intelligence-based software to perform frame-by-frame comparisons of the two videos. The software can quantify movement patterns and assist experts in assessing similarities between the recorded gait sequences.
As reported in Moneycontrol, police told the court,
We are planning to conduct a gait analysis of Chaudhary. We have CCTV footage showing him walking while wearing a hoodie to hide his face. The gait analysis will compare his locomotion in the CCTV footage with the recreated video to determine whether the walking patterns match.
Police Official
Forensic gait analysis is the scientific examination of an individual's walking pattern for investigative purposes. Rather than relying on facial features, the technique evaluates body movements that remain visible even when a person's face is concealed or the image quality is poor.1
Experts study several observable characteristics, including:
Stride and step length
Walking speed
Posture
Arm movement
Foot placement
Limb coordination
Body symmetry during walking
These features are compared with reference recordings to determine whether the movement patterns are consistent.2
The available CCTV footage reportedly did not capture the suspect's face clearly because it was covered by a hood. Although facial identification was not possible, the surveillance cameras recorded the individual's body movements throughout the walk.3
Although individuals often have recognizable walking characteristics, gait is not completely fixed. Several factors can temporarily alter the way a person walks, including:
Previous injuries or pain
Fatigue
Footwear
Carrying heavy objects
Uneven terrain
Age-related changes
Neurological or musculoskeletal disorders
Intentional disguise (such as limping deliberately or altering walking speed)
Because these variables may affect gait, forensic experts interpret gait findings together with other evidence collected during an investigation.1
Unlike fingerprints or DNA profiling, gait analysis is not considered a definitive biometric identifier and cannot independently establish a person's identity with complete certainty. Instead, it is primarily used as corroborative forensic evidence.¹²
Although gait analysis has been accepted as an investigative aid in several jurisdictions, experts caution that its reliability depends on the quality of the video footage, camera angle, lighting, clothing, walking surface, and the expertise of the examiner. It is generally regarded as supportive rather than standalone identification evidence.¹²
Artificial intelligence is becoming an important tool in forensic video analysis. AI-assisted software can examine surveillance footage frame by frame, extract measurable movement features, and compare walking characteristics with greater precision than visual observation alone. However, AI does not independently identify a suspect. Its findings are interpreted by trained forensic experts alongside other evidence.
Recent advances in computer vision and machine learning have improved movement analysis from surveillance videos, although expert review remains essential before drawing forensic conclusions.¹²
The technology is increasingly being used in investigations where facial recognition is not feasible because of face coverings, low-resolution recordings, or unfavorable camera angles.
As reported in NDTV, cyber expert Raj Shekhar said that recent advances in artificial intelligence have significantly improved the accuracy and effectiveness of forensic gait analysis.
The Pune police investigation remains ongoing. Officials have indicated that AI-assisted gait analysis will be evaluated alongside CCTV footage, digital evidence, crime scene reconstruction, and other forensic findings. Police are also seeking permission to conduct polygraph (lie detector) examinations of the accused as part of the investigation.
References
1. Badiye, Ashish, Neeti Kapoor, Prachi Kathane, and Kewal Krishan. "Forensic Gait Analysis." In StatPearls. Treasure Island, FL: StatPearls Publishing, 2020–. Last updated May 15, 2020. Accessed July 2, 2026. https://www.ncbi.nlm.nih.gov/books/NBK544363/.
2. Seckiner, Dilan, Xanthé Mallett, Philip Maynard, Didier Meuwly, and Claude Roux. "Forensic Gait Analysis—Morphometric Assessment from Surveillance Footage." Forensic Science International 296 (2019): 57–66. https://doi.org/10.1016/j.forsciint.2019.01.007
(Rh/SS/MSM)
