Investigation of Time-Frequency Feature Combinations with Histogram Layer Time Delay Neural Networks

TL;DR

This paper explores how different combinations of time-frequency features affect the performance of histogram layer time delay neural networks in underwater acoustic signal classification, highlighting the importance of feature engineering.

Contribution

It identifies optimal feature combinations for histogram layer TDLNs, demonstrating improved performance over single features in underwater acoustic signal analysis.

Findings

Certain feature combinations outperform single features

Optimal feature sets enhance neural network performance

Feature engineering remains crucial despite deep learning advances

Abstract

While deep learning has reduced the prevalence of manual feature extraction, transformation of data via feature engineering remains essential for improving model performance, particularly for underwater acoustic signals. The methods by which audio signals are converted into time-frequency representations and the subsequent handling of these spectrograms can significantly impact performance. This work demonstrates the performance impact of using different combinations of time-frequency features in a histogram layer time delay neural network. An optimal set of features is identified with results indicating that specific feature combinations outperform single data features.