Joseph the MoUSE — Mouse Ultrasonic Sound Explorer

Abstract

Joseph the MoUSE — Mouse Ultrasonic Sound Explorer (MoUSE) software aims to address the issue of manual analysis of recordings from experiments on rodents by introducing automatic techniques for ultrasonic vocalization (USV) detection. It combines deep learning (DL) methods with classical pattern recognition and computer graphics algorithms. During development, we used a dataset that consisted of recordings from real-world experiments in the open field. Recordings like these pose obstacles to automatic USV detection, one of which is the noise produced by mice in the experimental area or in nearby cages. Therefore, additionally, we conducted research and implemented de-noising methods along with detection algorithms. The project includes Python packages with algorithms for sound noise removal and USV detection, and provides a user-friendly graphical interface.

Keywords

• Object detection
• Image processing
• Ultrasonic vocalizations