Classifying Urban Sounds using Deep Learning
The goal of this project is to classify urban sounds into ten classes (jackhammer, engine_idling, siren, children_playing, drilling, street_music, air_conditioner, dog_bark, car_horn and gun_shot). The dataset is available for download at https://datahack.analyticsvidhya.com/contest/practice-problem-urban-sound-classification/. I have extracted the features from the audio files using the librosa library. I have implemented a Multilayer Perceptron (MLP) and a Convolutional Neural Network (CNN) model using tensorflow and keras for the classification task. The MLP model achieved a validation accuracy of over 90% whereas the CNN model achieved an accuracy of over 93% on the validation set.
P.S.: This is my first foray into audio/sound data.