Incremental Learning Algorithm For Sound Event Detection

Abstract

This paper presents a new learning strategy for the Sound Event Detection (SED) system to tackle the issues of i) knowledge migration from a pre-trained model to a new target model and ii) learning new sound events without forgetting the previously learned ones without re-training from scratch. In order to migrate the previously learned knowledge from the source model to the target one, a neural adapter is employed on the top of the source model. The source model and the target model are merged via this neural adapter layer. The neural adapter layer facilitates the target model to learn new sound events with minimal training data and maintaining the performance of the previously learned sound events similar to the source model. Our extensive analysis on the DCASE16 and US-SED dataset reveals the effectiveness of the proposed method in transferring knowledge between source and target models without introducing any performance degradation on the previously learned sound events while obtaining a competitive detection performance on the newly learned sound events.

Few-Shot Continual Learning for Audio Classification

“…Continual learning (never-ending learning, incremental learning, lifelong learning) [20][21][22], in contrast, is an online learning strategy where an algorithm seeks to continuously adapt to a sequence of tasks and perform well on all tasks without forgetting. It has been proposed for sound classification [23] and sound event detection [24] to learn new sound events without forgetting the previously learned ones. However, continual learning approaches typically require retraining when introducing novel classes, complicated training procedure, or large amounts of labeled data of the novel classes, which are not ideal for practical application with resourceconstrained computing environments or audio domains.…”

An Incremental Class-Learning Approach with Acoustic Novelty Detection for Acoustic Event Recognition

“…Most of ICL works have focused on computer vision tasks such as image classification [ 4 ], semantic segmentation [ 5 ], image classification in a number of isolated tasks [ 6 ]. Only a few [ 7 , 8 ] have focused on the incremental learning of new acoustic events for detection of the events. However, incremental learning without forgetting may also be useful for various tasks such as speech recognition, voice detection, acoustic scene analysis (ASA), acoustic event recognition (AER), acoustic anomaly detection (AAD), acoustic novelty detection (AND).…”

“…A few recent works have focused on incremental learning using CNN for AER. In [ 7 , 8 ], the performances of incremental learning were evaluated using Mel-spectrograms from one-second audio files. However, the detection of novel classes in the solutions to incremental learning, has not been addressed in the ICL studies.…”