SOUND EVENT DETECTION IN DOMESTIC ENVIRONMENTS USING ENSEMBLE OF CONVOLUTIONAL RECURRENT NEURAL NETWORKS Technical Report

In this paper, we present a method to detect sound events in domestic environments using small weakly labeled data, large unlabeled data, and strongly labeled synthetic data as proposed in the Detection and Classification of Acoustic Scenes and Events (DCASE) 2019 challenge task 4. To solve the problem, we use convolutional recurrent neural network (CRNN), as it stacks convolutional neural networks (CNN) and bi-directional gated recurrent unit (Bi-GRU). Moreover, we propose various methods such as data augmentation, event activity detection, multi-median filtering, mean-teacher student model, and the ensemble of neural networks to improve performance. By combining the proposed method, sound event detection performance can be enhanced, compared with the baseline algorithm. As a result, performance evaluation shows that the proposed method provides detection results of 40.89% for event-based metrics and 66.17% for segmentbased metrics.