Two Ways to Learn Audio Embeddings

Author(s): Edward Ma

Originally published on Towards AI.

Speech2Vec with Skip-gram and CBOW

Photo by Álvaro Bernal on Unsplash

Mel-frequency cepstral coefficients (MFCC), zero-crossing rate are some of classical feature for audio. It can be extracted via the library easily. However, it may not able to provide a high-quality signal or input for deep learning models nowadays.

Two teams of researchers propose a different way to learn audio embeddings but not leveraging those classical features. Chung and Glass (2018) proposes to learn word-based embeddings while Haque et al. (2019) suggests learning sentence-based embeddings.

Chung and Glass are inspired by word2vec to propose a different way to learn audio embeddings. word2vec leverages skip-gram or continuous bag-of-word (CBOW)… Read the full blog for free on Medium.

Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date with the latest developments in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming a sponsor.

Published via Towards AI