Convolutional Neural Networks with Multi-task Loss for Polyphonic Sound Event Detection
Huang Liu, Xiu Wang, Faqian Guan, Jinsen Hu · Proceedings of the 2nd International Conference on Computer Science and Application Engineering · 2018
The1 sound event detection methods which use convolution neural networks as classifiers have recently been demonstrated to improve the performance. Although the cross-entropy loss functions for classification used in general settings, it is controversial that loss functions shoul be tailored to specific tasks. Multi-task systems improve the generalization capability for some tasks. We use a multi-task loss function to couple with different neural networks and apply it to a polyphonic sound event detection task. The proposed method is compared with DNN, CNN and CBRNN methods. And experimental results show that the proposed method has made a considerable improvement for three different datasets.