Deep Learning based Detection and Delineation Sound in Smart City System

For many years, sound delineation has been a prominent area of research. Human speech is the basis for sound categorization models and sophisticated sound recognition systems features related to hearing. There are limitless applications for those traits, such as context awareness, criminal detection, and surveillance systems. In classifying audio dataset, it is proven to be highly effective convolutional neural network among the types of deep learning model. In order to do this, it describes sound files using the spectral contrast, chromogram, tonal centroid, and mel spectrogram as their picture representations, and then use these representations to train the deep learning model. The final outcome, especially considering the dataset’s characteristics and the fact that classifying ambient noises is considerably more difficult than classifying speech and music. The findings of the experiment likewise demonstrate a significant improvement in accuracy between the deep learning technique so that can be incorporated in surveillance system to delineation sound in crime environment in Smart City System.