Presentation Title: Gunshot Sound Detection Using Deep Learning Techniques

Presenter(s): Shinjini Gupta, Varsha Vaman Murthy, Sonal Yadav, & Sonam Barsainya, Department of Applied Data Science

Abstract: Usually, the gunshot events are mixed with some sort of background noise such as environmental sounds, human voice, animal sound, vehicle sound, and many more. It is important to come up with a solution that not only detects the gunshot sound from the audio recording but also filters out the background noise from the gunshot sound for better detection. So, we aim to propose a cost-effective gunshot detection solution using a Deep learning approach. For our project, we aim to utilize a publicly available dataset named the urbansound8k dataset. We will be using the Mel-Frequency Cepstral Coefficients (MFCC) feature of the audio sample and CNN model for classification. We will begin with sound files, extract the MFCC Feature using an open-source python Library named Librosa, and input the features into a CNN models such as Vgg-16, Inception-v3, ResNet to analyze the audio signals from the audio recordings and detect gunshots, and produce the prediction of whether the sound is a gunshot or not.

Link to Recorded Presentation: https://drive.google.com/file/d/1f9CtbaEwN47GLBYXzIxbYE9zzg9sC9QT/view?usp=sharing