Deep neural network 기반 오디오 표식을 위한 데이터 증강 방법 연구 = Study on data augmentation methods for deep neural network-based audio tagging

국문 초록 (Abstract)

본 논문에서는 DNN(Deep Neural Network) 기반 오디오 표식을 위한 데이터 증강 방법을 연구한다. 본 시스템에서는 오디오 신호를 멜-스펙트로그램으로 변환하여 오디오 표식을 위한 심층신경망의 입력으로 사용한다. 적은 수의 훈련 데이터를 사용하는 경우 발생하는 문제를 해결하기 위해, 타임 스트레칭, 피치 변화, 동적 영역 압축, 블록 혼합 등의 방법을 사용하여 훈련 데이터를 증강시켰다. 사용된 데이터 증강 기법의 최적 파라미터와 최적 조합을 오디오 표식 시뮬레이션을 통해 확인하였다.

다국어 초록 (Multilingual Abstract)

In this paper, we present a study on data augmentation methods for DNN (Deep Neural Network)-based audio tagging. In this system, an audio signal is converted into a mel-spectrogram and used as an input to the DNN for audio tagging. To cope with the problem associated with a small number of training data, we augment the training samples using time stretching, pitch shifting, dynamic range compression, and block mixing. In this paper, we derive optimal parameters and combinations for the augmentation methods through audio tagging simulations.

참고문헌 (Reference)

1 Dolby E, "Standards and practices for authoring Dolby Digital and Dolby E bitstreams" Dolby Labortories, Inc 2002

2 S. M. Beitzel, "On Understanding And Classifying Web Queries" Illinois Institute of Technology 2006

3 R. Seizel, "Large-scale weakly labeled semi-supervised sound event detection"

4 G. E. Dahl, "Improving DNNs for LVCSR using rectified linear units and dropout" 8609-8613, 2013

5 S. Mum, "Generative adversarial network based acoustic scene training set augmentation and selection using SVM hyper-plane" 93-97, 2017

6 D. Giannoulis, "Detection and classification of acoustic scenes and events: an IEEE AASP challenge" 1-4, 2013

7 J. Salamon, "Deep convolutional neural networks and data augmentation for environmental sound classification" 24 : 279-283, 2016

8 P. Cano, "Content-based music audio recommendation" 211-212, 2005

9 E. Wold, "Content-based classification, search, and retrieval of audio" 3 : 27-36, 1996

10 P. Foster, "CHiME-home: A dataset for sound source recognition in a domestic environment" 15 : 2015

11 M. Schuster, "Bidirectional recurrent neural networks" 45 : 2673-2681, 1997

12 J. F. Gemmeke, "Audio set: An ontology and human-labeled dataset for audio events" 776-780, 2017

13 M. Hilsamer, "A statistical approach to automated offline dynamic processing in the audio mastering process" 35-40, 2014