# Sound source detection, localization and classification using consecutive ensemble of CRNN models

## TL;DR

This paper presents a novel ensemble of CRNN models for sound event localization and detection, effectively estimating active sources, their directions, and classes, demonstrating competitive performance on a benchmark dataset.

## Contribution

It introduces a consecutive ensemble approach with four CRNN models to improve sound source detection, localization, and classification in a unified framework.

## Key findings

- Achieved accurate localization and classification on TAU dataset
- Outperformed some existing methods in DCASE2019 task3
- Demonstrated the effectiveness of consecutive ensemble models

## Abstract

In this paper, we describe our method for DCASE2019 task3: Sound Event Localization and Detection (SELD). We use four CRNN SELDnet-like single output models which run in a consecutive manner to recover all possible information of occurring events. We decompose the SELD task into estimating number of active sources, estimating direction of arrival of a single source, estimating direction of arrival of the second source where the direction of the first one is known and a multi-label classification task. We use custom consecutive ensemble to predict events' onset, offset, direction of arrival and class. The proposed approach is evaluated on the TAU Spatial Sound Events 2019 - Ambisonic and it is compared with other participants' submissions.