Audio Event Detection using Weakly Labeled Data

TL;DR

This paper introduces a novel framework for acoustic event detection using only weakly labeled data, formulating it as a multiple instance learning problem and employing SVMs and neural networks to detect and temporally locate events.

Contribution

It presents the first approach to acoustic event detection with weak labels by formulating it as multiple instance learning and proposing two effective frameworks.

Findings

Supports weakly labeled data for event detection

Provides temporal localization of events

Reduces need for manual annotation

Abstract

Acoustic event detection is essential for content analysis and description of multimedia recordings. The majority of current literature on the topic learns the detectors through fully-supervised techniques employing strongly labeled data. However, the labels available for majority of multimedia data are generally weak and do not provide sufficient detail for such methods to be employed. In this paper we propose a framework for learning acoustic event detectors using only weakly labeled data. We first show that audio event detection using weak labels can be formulated as an Multiple Instance Learning problem. We then suggest two frameworks for solving multiple-instance learning, one based on support vector machines, and the other on neural networks. The proposed methods can help in removing the time consuming and expensive process of manually annotating data to facilitate fully supervised learning. Moreover, it can not only detect events in a recording but can also provide temporal locations of events in the recording. This helps in obtaining a complete description of the recording and is notable since temporal information was never known in the first place in weakly labeled data.