Recurrent Neural Networks with Stochastic Layers for Acoustic Novelty Detection
This addresses the problem of detecting novel acoustic events for applications like surveillance or monitoring, but it is incremental as it adapts an existing method to a new task.
The paper tackled acoustic novelty detection by adapting Recurrent Neural Networks with Stochastic Layers to learn distributions of complex sequences, and it outperformed state-of-the-art detectors on a benchmark dataset.
In this paper, we adapt Recurrent Neural Networks with Stochastic Layers, which are the state-of-the-art for generating text, music and speech, to the problem of acoustic novelty detection. By integrating uncertainty into the hidden states, this type of network is able to learn the distribution of complex sequences. Because the learned distribution can be calculated explicitly in terms of probability, we can evaluate how likely an observation is then detect low-probability events as novel. The model is robust, highly unsupervised, end-to-end and requires minimum preprocessing, feature engineering or hyperparameter tuning. An experiment on a benchmark dataset shows that our model outperforms the state-of-the-art acoustic novelty detectors.