Forensic Physics: Voice, Video and Reconstruction

Third and final hard premium FACT Forensic Physics mock — closes the series. Coverage of four FACT Forensic Physics syllabus sub-topics: math & statistics (likelihood ratios, prosecutor's fallacy, twin DNA, confidence intervals, regression interpretation, Bayes Nets, Type I/II errors), forensic voice authentication (spectrogram interpretation, F0 disguise detection, dialect-aware comparison, ASR vs auditory-acoustic methods, deepfake-voice detection, voice-morphing artefacts), forensic video analysis (H.264 frame extraction, video authenticity, photogrammetric height reconstruction, super-resolution and AI hallucination, PRNU camera fingerprinting, frame interpolation as visualisation), and collision investigation & reconstruction (pedestrian-throw distance, EDR pre-crash data, momentum conservation in multi-vehicle collisions, yaw-mark speed estimation, autonomous-vehicle TTC analysis, helmet IS 4151 testing, airbag-without-seatbelt, breath-blood alcohol partition coefficient, night-time visibility and unlit-vehicle responsibility, sensor degradation under fog).

It is pitched at advanced MSc forensic-science students at NFSU, GFSU, LNJN-NICFS and other Indian universities, FACT and FACT Plus aspirants in their final preparation phase, and UGC-NET candidates calibrating their breadth across forensic statistics, voice, video, and collision physics. The questions push toward contemporary applications: deepfake-voice detection, AV collision investigation, super-resolution forensics, and isotope-dilution toxicology. Pair with #6 (Foundations), #7 (Applied Analysis), #8 (Evidence Collection & Pattern Analysis), and #9 (Instruments & Spectroscopy) for the complete five-paper FACT Forensic Physics series.

Topics covered:

• Likelihood ratio interpretation, prosecutor's fallacy, verbal scales (RSS / ENFSI)
• Identical-twin STR profile sharing and post-zygotic mutation detection
• Confidence interval frequentist interpretation and lower-bound conservative reporting
• r² vs p-value vs causation vs individualisation distinctions
• Bayesian Networks for multi-evidence dependent inference
• Type I (false positive, α) vs Type II (false negative, β) errors
• Glass random match probability ≈ 1 in 10,000 from elements + n
• Spectrogram interpretation: F0 + formants for speaker classification
• Audio enhancement (spectral subtraction, Wiener) discipline
• Dialect/sociolect/idiolect impact on speaker comparison
• ASR (x-vector / GMM-UBM) vs AAP analysis combination
• Deepfake-voice detection: breathing, prosody, ASVspoof
• Pitch-shift disguise detection (formants don't shift naturally)
• H.264 / H.265 frame-type extraction (I, P, B frames)
• Video authenticity: timestamps + artefacts + sync + hash + PRNU
• Photogrammetric height reconstruction (h × D = constant)
• Super-resolution: legitimate vs AI hallucination
• PRNU sensor-noise fingerprinting for camera identification
• Pedestrian-throw distance equations (Searle / Han / Wood)
• EDR pre-crash data interpretation (5-15 second window)
• Conservation of momentum for multi-vehicle inelastic collisions
• Yaw-mark radius and v = √(g × r × μ)
• Autonomous-vehicle TTC + reaction + max deceleration analysis
• Helmet IS 4151 forensic testing protocol
• Airbag effectiveness conditional on seatbelt use
• Breath-blood alcohol Henry's law and 2100:1 partition coefficient
• Night-time visibility geometry and unlit-vehicle responsibility
• Frame interpolation: visualisation only, not evidence
• AV sensor degradation in fog: camera vs lidar vs radar

Each question carries a 220+ word structured explanation citing standard references (Aitken & Taroni statistics, Butler DNA typing, Curran statistics, Hollien voice ID, ENFSI guidelines, Rose forensic speaker recognition, ASVspoof challenge, SWGDE video forensics, Lukáš PRNU, Searle pedestrian-throw, SAE J1698 EDR, Daily & Strickland collision reconstruction, Olson & Sivak perception-reaction, Bureau of Indian Standards IS 4151 helmets, NHTSA occupant protection, ICADTS alcohol). Allow 30 minutes; explanations double as study notes for the contemporary-applications paper. This mock completes the FACT Forensic Physics five-paper series; together with Mocks #6-#10, the entire syllabus sub-section is covered at three difficulty levels.