Deep Learning Based Speech Quality Prediction by Gabriel Mittag

Author:Gabriel Mittag
Language: eng
Format: epub
ISBN: 9783030914790
Publisher: Springer International Publishing

Cosine

Another function that is often used is the cosine similarity (Graves et al. 2014), which can be seen as a variation of the dot product that neglects the magnitudes of the vectors. In case of normalised vectors, the cosine similarity delivers the same results as the dot product. It is calculated as follows:

(4.2)

where 𝜖 is set to 1e-08 to avoid zero division.

Luong

This alignment score function is similar to the dot product, however, before calculating the dot product between the two feature vectors, the reference signal is multiplied with a trainable weight matrix according to Luong et al. (2015) as follows:

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.