Echoic log-surprise: A multi-scale scheme for acoustic saliency detection

作者：

Highlights：

• A new model to compute acoustic saliency based on Bayesian log-surprise is proposed.

• Our model implements a multi-scale scheme, inspired by Acoustic Sensory Memory.

• We validated our proposal using Acoustic Event Detection datasets.

• Performance compared favorably against state-of-the-art saliency algorithms.

摘要

•A new model to compute acoustic saliency based on Bayesian log-surprise is proposed.•Our model implements a multi-scale scheme, inspired by Acoustic Sensory Memory.•We validated our proposal using Acoustic Event Detection datasets.•Performance compared favorably against state-of-the-art saliency algorithms.

论文关键词：Acoustic saliency,Echoic memory,Multi-scale,Statistical divergence,Jensen–Shannon,Acoustic Event Detection

论文评审过程：Received 30 November 2017, Revised 6 July 2018, Accepted 7 July 2018, Available online 20 July 2018, Version of Record 3 August 2018.

论文官网地址：https://doi.org/10.1016/j.eswa.2018.07.018