A Review on Scream Detection and Classification

A Review on Scream Detection and Classification

Harsh Pandey, Jai Rai Lakhi, Vaishali Goel, Deepika Bansal

Dept. of Information Technology
Maharaja Agrasen Institute of Technology
Delhi, India

Abstract—In contemporary living environments, the examination of non-speech audio signals is vital for enhancing situational awareness, complementing visual information from video signals. Screams within non-speech audio signals are particularly important for individuals like security personnel, caregivers, and family members due to their association with potential danger. Contrary to traditional beliefs, this review concentrates on automated acoustic systems specialized in non-speech scream classification. The objective is to show that screams can be categorized into emotional classes such as happiness, sadness, fear, and danger. Drawing inspiration from advancements in scream audio detection and classification, the review provides a taxonomy detailing target applications, prominent sound features, classification techniques, and their evolution over recent decades. This comprehensive review aims to assist researchers in choosing the most suitable scream detection and classification techniques, along with relevant acoustic parameters, contributing to a deeper understanding of speaker vocalization conditions.

Keywords— Scream detection, scream classification, classification techniques, non-speech audio signals.