Family Conflict Intensity Detection Analysis

Problem

Fixed living-room cameras often capture what happened without quickly indicating whether a family dispute has moved beyond ordinary disagreement. High TV volume, children playing, or heated discussion can be misread, while repeated high-intensity conflict may affect mental health and escalate. This skill frames the problem as extracting objective acoustic and visual indicators from mic-enabled video, with aggressive-word matching handled locally, to help assess conflict intensity.

How it works

• Input: mp4/avi/mov files with an audio track, preferably covering at least 30 seconds of the full interaction and showing upper bodies; audio at ≥16 kHz is recommended.
• Acoustic metrics: peak_db, avg_db, db_delta_vs_baseline, shout_event_count, aggressive_word_hit_count, voice_speakers_estimate.
• Visual metrics: wave_hand_event_count, finger_point_event_count, push_event_count, throw_object_event_count, physical_proximity_invasion.
• Combined assessment: outputs low / medium / high, a gentle_reminder_message, and recommended actions such as cooling down, separating rooms, or optional emergency-contact notification.
• History reports: uses --list to query cloud reports and renders them as a Markdown table with report name, intensity/main behavior, analysis time, and a link.

Boundaries

It only provides objective intensity metrics and gentle reminders, not legal advice, psychological treatment, or safety determinations. Emergency-contact linkage must be enabled with prior consent from both parties. For repeated high-intensity conflict or safety concerns, contact the anti-domestic-violence hotline 12338 or local police and mental-health services. Raw home video is highly sensitive; obtain consent from all adult household members and prefer silhouette mode, facial mosaic, and storing only metric statistics.