An Explainable Proxy Model for Multiabel Audio Segmentation