Detection and Localization of Audio Event for Home Surveillance Using CRNN

Safety and security have been a prime priority in people’s lives, having surveillance system at home keeps people their property more secured. In this paper, an audio has proposed that does both the detection localization of or sound events. The combined task detecting localizing events is known as Sound Event Localization (SELD). SELD work executed through Convolutional Recurrent Neural Network (CRNN) architecture. CRNN stacked layer convolutional neural network (CNN), recurrent (RNN) fully connected (FNN). takes multichannel input, extracts features input parallel. results obtained by with gated unit (GRU) long short-term memory (LSTM) are compared discussed paper. LSTM gives 75% F1 score 82.8% frame recall for one overlapping sound. Therefore, uses produces better overall performance