Spectrogram features for audio and speech analysis

TL;DR

This paper reviews how spectrogram features are used in audio and speech analysis, examining their properties, variations, and how they interact with classifier architectures across different tasks.

Contribution

It provides a comprehensive survey of spectrogram-based representations, analyzing their characteristics and their compatibility with various machine learning models for audio analysis.

Findings

Spectrogram parameters significantly affect analysis performance.

Different spectrogram configurations suit different audio tasks.

The choice of spectrogram features influences classifier effectiveness.

Abstract

Spectrogram-based representations have grown to dominate the feature space for deep learning audio analysis systems, and are often adopted for speech analysis also. Initially, the primary motivator for spectrogram-based representations was their ability to present sound as a two dimensional signal in the time-frequency plane, which not only provides an interpretable physical basis for analysing sound, but also unlocks the use of a wide range of machine learning techniques such as convolutional neural networks, that had been developed for image processing. A spectrogram is a matrix characterised by the resolution and span of its two dimensions, as well as by the representation and scaling of each element. Many possibilities for these three characteristics have been explored by researchers across numerous application areas, with different settings showing affinity for various tasks. This paper reviews the use of spectrogram-based representations and surveys the state-of-the-art to question how front-end feature representation choice allies with back-end classifier architecture for different tasks.