Overlapping acoustic event classification based on joint training with source separation

Soo Hyun Bae, In Kyu Choi, Hyung Yong Kim, Kang Hyun Lee, Nam Soo Kim

Abstract: Overlapping acoustic event classification is the task of estimating multiple acoustic events in a mixed source. In the case of non-overlapping event classification, many approaches have achieved a great success using various feature extraction methods and deep learning models. However, in most real life situations, acoustic events are overlapped and different events may share similar properties. Simultaneously detecting mixed sources is a challenging problem. In this paper, we propose a classification method for overlapping acoustic events which incorporates joint training with source separation framework. Since overlapping acoustic events are mixed in multiple sources, we trained the source separation model and multi-label classification model for estimating the type of overlapping acoustic events. The source separation model is trained to reconstruct the target sources by minimizing the interference of overlapping events. Joint training can be conducted to achieve end-to-end optimization between the acoustic event source separation and multi-label estimation. To evaluate the proposed method, we conducted a number of experiments using artificially mixed data. We observed that the jointly trained neural network outperforms the baseline network with an identical structure except for the training method.