Environmental Sounds Recognition with Convolutional-LSTM

This paper addresses the task of recognizing environmental sounds using the AudioSet data set. Specifically, features were extracted by spectrogram conversion of AudioSet's 10-second sound data, and learning was performed by neural network with 3 layer CNN and the one layer CLSTM, and a comparison experiment with the LSTM model was performed. As a result, CLSTM increased 4 points from 52% to 56% better recognition accuracy than LSTM model, demonstrating the usefulness of CLSTM in environmental sound recognition field.