Self-supervised learning for Environmental Sound Classification

Tripathi, Achyut Mani; Mishra, Aakansha

doi:10.1016/j.apacoust.2021.108183

Cited by 33 publications

(6 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…However, when the model performance is compared to the state-of-the-art performances achieved for the broader ESC domain, it can be identified that the results published in this paper require further refinement. Therefore, as future directions, applying transfer learning using the ImageNet dataset [ 36 , 64 , 65 ], exploring different data augmentation techniques [ 32 , 46 , 66 ] and feature representation methodologies [ 67 , 68 ] are suggested by the authors.…”

Section: Discussionmentioning

confidence: 99%

Forest Sound Classification Dataset: FSC22

Bandara

Jayasundara

Ariyarathne

et al. 2023

Sensors

View full text Add to dashboard Cite

The study of environmental sound classification (ESC) has become popular over the years due to the intricate nature of environmental sounds and the evolution of deep learning (DL) techniques. Forest ESC is one use case of ESC, which has been widely experimented with recently to identify illegal activities inside a forest. However, at present, there is a limitation of public datasets specific to all the possible sounds in a forest environment. Most of the existing experiments have been done using generic environment sound datasets such as ESC-50, U8K, and FSD50K. Importantly, in DL-based sound classification, the lack of quality data can cause misguided information, and the predictions obtained remain questionable. Hence, there is a requirement for a well-defined benchmark forest environment sound dataset. This paper proposes FSC22, which fills the gap of a benchmark dataset for forest environmental sound classification. It includes 2025 sound clips under 27 acoustic classes, which contain possible sounds in a forest environment. We discuss the procedure of dataset preparation and validate it through different baseline sound classification models. Additionally, it provides an analysis of the new dataset compared to other available datasets. Therefore, this dataset can be used by researchers and developers who are working on forest observatory tasks.

show abstract

Section: Discussionmentioning

confidence: 99%

Forest Sound Classification Dataset: FSC22

Bandara

Jayasundara

Ariyarathne

et al. 2023

Sensors

View full text Add to dashboard Cite

show abstract

“…The initial evaluation involved assessing the enhancement of SubSpectralNet [6] 74.08 / MCTA-CNN [7] 72.40 75.71 Atrous-CNN [8] 72.7 / ResNet [11] / 75.1 [12] 72.7 72.2 CompactNet [13] / 69.0 SubSpectralNet [6] 74.1 / CMAM [14] / 68.8 Log-Mel-DNN [15] 68.2 66.8…”

Section: Methodsmentioning

confidence: 99%

A CNN-Based Multi-Scale Pooling Strategy for Acoustic Scene Classification

HUANG,

XIE

2024

IEICE Trans. Inf. & Syst.

View full text Add to dashboard Cite

Acoustic scene classification (ASC) is a fundamental domain within the realm of artificial intelligence classification tasks. ASCbased tasks commonly employ models based on convolutional neural networks (CNNs) that utilize log-Mel spectrograms as input for gathering acoustic features. In this paper, we designed a CNN-based multi-scale pooling (MSP) strategy for ASC. The log-Mel spectrograms are utilized as the input to CNN, which is partitioned into four frequency axis segments. Furthermore, we devised four CNN channels to acquire inputs from distinct frequency ranges. The high-level features extracted from outputs in various frequency bands are integrated through frequency pyramid average pooling layers at multiple levels. Subsequently, a softmax classifier is employed to classify different scenes. Our study demonstrates that the implementation of our designed model leads to a significant enhancement in the model's performance, as evidenced by the testing of two acoustic datasets.

show abstract

“…A signal is a variation of variables such as the pressure wave of sound, the color of an image, the depth of a surface, the temperature of a body, the voltage or current of a conductor or biological system, light, radio electromagnetic signals, the price of goods or the volume and weight of an object. It can be said that a signal is a medium to carry information about the past, present and future state of a variable [17].…”

Section: B Animal Soundmentioning

confidence: 99%

Sound Classification for Javanese Eagle Based on Improved Mel-Frequency Cepstral Coefficients and Deep Convolutional Neural Network

Permana,

Rahman

2024

IJACSA

View full text Add to dashboard Cite

The Javanese Eagle is a rare and protected animal in Indonesia. These animals only live in a few species and are threatened with extinction. These birds need to be bred to avoid extinction. One form of communication between the Javanese eagles and each other is the sound of their tweets. These tweets can be studied and classified to conserve endangered animals. This study will classify the sound of the Javanese Eagles for the benefit of animal conservation. Data in the form of voice tweets will be classified. This classification uses algorithms from improved MFCC (Mel-Frequency Cepstral Coefficients) and Deep Convolutional Neural Network. The result of this study was to classify the sound of the Javanese Eagle from the lack of food or drink, the normal tweets state of the bird, and to find out the Javanese Eagle in finding a partner. This research has been carried out by comparing the CNN architecture with AlexNet and VGGNet models and various combinations of training, validation, and test data. The best model dataset underwent division into 80% for training, 10% for validation, and 10% for testing. Training and testing of both IMFCC and VGGNet models occurred using the same dataset. During training, VGGNet achieved 100% accuracy, while testing yielded 99%. ROC Curve: 'Normal' AUC 0.996, 'Looking for Partner' AUC 1.000, 'Looking for Food' AUC 0.996. This study aids Javanese Eagle conservation, crucial for preventing extinction at conservation sites.

show abstract

Self-supervised learning for Environmental Sound Classification

Cited by 33 publications

References 27 publications

Forest Sound Classification Dataset: FSC22

Forest Sound Classification Dataset: FSC22

A CNN-Based Multi-Scale Pooling Strategy for Acoustic Scene Classification

Sound Classification for Javanese Eagle Based on Improved Mel-Frequency Cepstral Coefficients and Deep Convolutional Neural Network

Contact Info

Product

Resources

About