Diverse Audio Embeddings -- Bringing Features Back Outperforms CLAP!

TL;DR

This paper demonstrates that combining diverse handcrafted audio feature embeddings with end-to-end learned representations significantly improves sound classification performance over using end-to-end models alone.

Contribution

It introduces a method to integrate domain-specific handcrafted audio embeddings with end-to-end models, achieving superior classification results.

Findings

Handcrafted embeddings alone do not outperform end-to-end models.

Combining handcrafted and end-to-end embeddings improves accuracy.

The approach surpasses traditional end-to-end training in audio classification.

Abstract

With the advent of modern AI architectures, a shift has happened towards end-to-end architectures. This pivot has led to neural architectures being trained without domain-specific biases/knowledge, optimized according to the task. We in this paper, learn audio embeddings via diverse feature representations, in this case, domain-specific. For the case of audio classification over hundreds of categories of sound, we learn robust separate embeddings for diverse audio properties such as pitch, timbre, and neural representation, along with also learning it via an end-to-end architecture. We observe handcrafted embeddings, e.g., pitch and timbre-based, although on their own, are not able to beat a fully end-to-end representation, yet adding these together with end-to-end embedding helps us, significantly improve performance. This work would pave the way to bring some domain expertise with end-to-end models to learn robust, diverse representations, surpassing the performance of just training end-to-end models.