Environmental sound classification using convolution neural networks with different integrated loss functions

Abstract

The hike in the demand for smart cities has gathered the interest of researchers to work on environmental sound classification. Most researchers' goal is to reach the Bayesian optimal error in the field of audio classification. Nonetheless, it is very baffling to interpret meaning from a three-dimensional audio and this is where different types of spectrograms become effective. Using benchmark spectral features such as mel frequency cepstral coefficients (MFCCs), chromagram, log-mel spectrogram (LM), and so on audio can be converted into meaningful 2D spectrograms. In this paper, we propose a convolutional neural network (CNN) model, which is fabricated with additive angular margin loss (AAML), large margin cosine loss (LMCL) and a-softmax loss. These loss functions proposed for face recognition, hold their value in the other fields of study if they are implemented in a systematic manner. The mentioned loss functions are more dominant than conventional softmax loss when it comes to classification task because of its capability to increase intra-class compactness and interclass discrepancy. Thus, with MCAAM-Net, MCAS-Net and MCLCM-Net models, a classification accuracy of 99.60%, 99.43% and 99.37% is achieved on UrbanSound8K dataset respectively without any augmentation. This paper also demonstrates the benefit of stacking features together and the above-mentioned validation accuracies are achieved after stacking MFCCs and chromagram on the x-axis. We also visualized the clusters formed by the embedded vectors of test data for further acknowledgement of our results, after passing it through different proposed models. Finally, we show that the MCAAM-Net model achieved an accuracy of 99.60% on UrbanSound8K dataset, which outperforms the benchmark models like TSCNN-DS, ADCNN-5, ESResNet-Attention, and so on that are introduced over the recent years.additive angular margin, angular softmax, chromagram, large margin cosine, mel frequency cepstral coefficient, smart city | INTRODUCTIONOver the years, computers are given tortuous work in the field of deep learning (DL) and surprisingly computers did not disappoint, even subduing human capabilities at times. Along with advancement in the field of computer vision (CV), analysis of audio recognition has also being thriving (Dang et al., 2020). However, speech recognition and music processing has been the focal points in the province of audio analysis (Piczak, 2015b).But in the last few years, environmental sound classification (ESC) have been in the spotlight with the growth of the smart city applications and

Analysis and Acoustic Event Classification of Environmental Data Collected in a Citizen Science Project

“…Garg et al worked with the same database using MFCC along with a CNN and a Long Short-Term Memory (LSTM) with test data accuracies ranging from 77% to 82% [ 50 ]. Das et al complemented the CNN with different integrated loss functions achieving even higher accuracies on the same dataset [ 51 ]. The same author also tested CNNs and LSTMs combined with up to seven different features, including MFCC and diverse Chroma features with state-of-the-art performance [ 52 ].…”

Transformers for Urban Sound Classification—A Comprehensive Performance Evaluation

“…Like the image classification problem, a CNN is a natural architecture for audio classification. Therefore, researchers explored the use of this architecture, namely: Salamon and Bello [23], by employing a Deep Convolutional Neural Network (DCNN); Das et al [3] explored the use of a CNN model with a specific Additive Angular Margin Loss (AAML) and also explored a CNN combined with stacked features such as Mel Frequency Cepstral Coefficients (MFCC) and Chromagram; Mu et al [6] introduced the Temporal-Frequency Attention-Based Convolutional Neural Network (TFCNN) model, a CNN-based model associated with attention mechanisms, among others.…”

“…Thus, the scientific community has been developing different computational algorithms to acquire, analyse, and classify urban sounds automatically. Nonetheless, the combination of multiple classes, abnormal noise conditions, and the multiplicity of sound sources are still limitations for efficiently completing the task [1,[3][4][5].…”

Sound Classification and Processing of Urban Environments: A Systematic Literature Review

“…Recently, smart cities are emerging to take advantage of all opportunities that cities can provide to improve the lives of their citizens, such as taking advantage of the sensing architecture spread around the city to create innovative services (Bello et al [2]). Accordingly, one of the main requirements concerns Urban Sound characterization, which encompasses several tasks, such as sound classification and segmentation, and still poses different challenges (Mushtaq and Su [3], Das et al [4]). It is estimated that major cities must handle thousands of co-occurring events, with rapid events that require immediate action passing unnoticed by authorities (Mushtaq and Su [3], Das et al [4]).…”

“…Accordingly, one of the main requirements concerns Urban Sound characterization, which encompasses several tasks, such as sound classification and segmentation, and still poses different challenges (Mushtaq and Su [3], Das et al [4]). It is estimated that major cities must handle thousands of co-occurring events, with rapid events that require immediate action passing unnoticed by authorities (Mushtaq and Su [3], Das et al [4]).…”