Mel-spectrogram and Deep CNN Based Representation Learning from Bio-Sonar Implementation on UAVs

Tanveer, M. Hassan; Zhu, Hongxiao; Ahmed, Waqar; Thomas, Antony; Imran, Basit Muhammad

doi:10.1109/icccr49711.2021.9349416

Cited by 19 publications

(12 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Here, we compared with another relevant state-of-the-art method that is based on scattergram and deep CNN (Convolutional Neural Network) [ 28 ]. The scattergram of size

was computed using WST and is similar to the mel-spectrogram when considering the filter bank 1 or layer 1 to compute the WST for finding the scattergram.…”

Section: Resultsmentioning

confidence: 99%

A New Acoustical Autonomous Method for Identifying Endangered Whale Calls: A Case Study of Blue Whale and Fin Whale

Sattar

2023

Sensors

View full text Add to dashboard Cite

In this paper, we study to improve acoustical methods to identify endangered whale calls with emphasis on the blue whale (Balaenoptera musculus) and fin whale (Balaenoptera physalus). A promising method using wavelet scattering transform and deep learning is proposed here to detect/classify the whale calls quite precisely in the increasingly noisy ocean with a small dataset. The performances shown in terms of classification accuracy (>97%) demonstrate the efficiency of the proposed method which outperforms the relevant state-of-the-art methods. In this way, passive acoustic technology can be enhanced to monitor endangered whale calls. Efficient tracking of their numbers, migration paths and habitat become vital to whale conservation by lowering the number of preventable injuries and deaths while making progress in their recovery.

show abstract

“…Here, we compared with another relevant state-of-the-art method that is based on scattergram and deep CNN (Convolutional Neural Network) [ 28 ]. The scattergram of size

was computed using WST and is similar to the mel-spectrogram when considering the filter bank 1 or layer 1 to compute the WST for finding the scattergram.…”

Section: Resultsmentioning

confidence: 99%

A New Acoustical Autonomous Method for Identifying Endangered Whale Calls: A Case Study of Blue Whale and Fin Whale

Sattar

2023

Sensors

View full text Add to dashboard Cite

show abstract

“…In the other hand, in Mel spectrogram, the frequency of the signal would be converted to log scale and amplitudes which could show the spectrogram. Afterward, the frequency domain would be mapped to the Mel scale to shape the Mel spectrogram at last [49,50].…”

Section: Definitions and Notationsmentioning

confidence: 99%

An LSTM-Autoencoder Architecture for Anomaly Detection Applied on Compressors Audio Data

Mobtahej

Zhang

Hamidi

et al. 2022

Computational and Mathematical Methods

View full text Add to dashboard Cite

The compressors used in today’s natural gas production industry have an essential role in maintaining the production line operational. Each of the compressors’ components has routine maintenance tasks to avoid sudden failures. Hence, the significant advantages and benefits of performing preventative maintenance tasks in time are decreasing equipment downtime, saving additional costs, and improving the safety and reliability of the whole system. In this paper, anomaly classification and detection methods based on a neural network hybrid model named Long Short-Term Memory (LSTM)-Autoencoder (AE) is proposed to detect anomalies in sequence pattern of audio data, collected by multiple sound sensors deployed at different components of each compressor system for predictive maintenance. In research methodology, this paper has conducted experiments that employed different RNN architectures such as GRU, LSTM, Stacked LSTM, and Stacked GRU with various functions to create a baseline for model evaluation. Each architecture used audio signals dataset received from the compressor system for experiments to consider each neural network model’s performance. According to performance results, an optimal model for anomaly detection with the best performance scores has been proposed in this research. Experiments combined one-dimensional raw audio signal features using SC and Mel spectrogram features were fed to deep learning models to evaluate performance. Hence, such hybrid methods can effectively detect normal and anomaly audio signals collected from a compressor system, increasing the compressor system’s reliability and the sustainability of the gas production line. The combination of multiple-resource features in the proposed hybrid model showed a 100% score in all four-evaluation metrics such as accuracy, precision, recall, and F1 in LSTM-based autoencoder in both test and train results.

show abstract

“…They are produced by generating Fourier transforms of a signal on short time windows and calculating the energy contained in each logarithmically-spaced frequency interval (known as a "Mel"). This technique has yielded excellent results in state-of-the-art deep learning audio classification studies [Hershey et al 2017] [Thornton [n.d.]] [Tanveer et al 2021] due to its favourable representation of the time-frequency attributes of a signal. Because of the logarithmic frequency scale, an increased or decreased fundamental frequency simply results in vertical translation on the plot without the dilation or contraction associated with typical spectrograms.…”

Section: Data Processing and Feature Extractionmentioning

confidence: 99%

Deep Learning for Enhanced Scratch Input

Bhargava¹,

Zhou²,

Carnaffan³

et al. 2021

Preprint

View full text Add to dashboard Cite

Mel-spectrogram and Deep CNN Based Representation Learning from Bio-Sonar Implementation on UAVs

Cited by 19 publications

References 24 publications

A New Acoustical Autonomous Method for Identifying Endangered Whale Calls: A Case Study of Blue Whale and Fin Whale

A New Acoustical Autonomous Method for Identifying Endangered Whale Calls: A Case Study of Blue Whale and Fin Whale

An LSTM-Autoencoder Architecture for Anomaly Detection Applied on Compressors Audio Data

Deep Learning for Enhanced Scratch Input

Contact Info

Product

Resources

About