ChiBa—A Chirrup and Bark Detection System for Urban Environment

The World is developing at a tremendous pace which has been catapulted by large-scale technological advancements. Building mega structures has never been easier and modes of commute have also developed thereby shortening travel-time. Such advancements have also brought along newer sources of pollution which are harming our planet at an even faster pace. Sound pollution is one such agent that has a long-term effect on not only humans but the entire biodiversity. Its effect on life is not immediately observed but the damage becomes visible over time. Birds are one of the most affected creatures due to sound pollution. This is one of the major reasons for declining bird population in the Urban areas. It is very important to preserve biodiversity for a sustainable future. Animals have calls that are melodious and rhythmic and these calls tend to change when they are in distress. An automated system can be very useful in this context which can monitor animal sounds and detect changes in their calls. Deployment of such a system in Urban areas is challenging due to the presence of ambient sounds which is extremely diverse. Thus it is essential to initially detect animal calls in the Urban environment prior to monitoring them. ChiBa is a system proposed to address this problem. Experiments were initially performed with the detection of birds and dogs (the most common and loudest creatures in cities) calls in the Urban environment. Tests were performed with over 7K clips comprising of the animal calls as well as Urban ambient sounds. The audios were modeled using a deep learning-based approach wherein the highest accuracy of 99.91% was obtained.