AI-Based Acoustic Surveillance System for Smart Cities

This paper provides an insight into current capabilities of artificial intelligence-based acoustic applications. The growth of machine learning, improvements in the dynamics and linearity of miniature microphones, and increased computational power of signal processors enable the development and deployment of applications that are capable of real-time, human-operator-free analysis of sounds in the environment and determination of selected phenomena, including their detailed identification and localization of the place of origin. As a detailed example described in the paper, a system for detecting, identifying, and localizing gunshot sources in public and indoors is presented. Being designed within the Internet of Things concept utilizing low-power sensors and AI-based signal processing, it provides information about the likely type of weapon and ammunition used, in addition to source localization for Smart-City infrastructures. The system ethical implications are also briefly discussed.