search

WeidiXie / VGG-Speaker-Recognition

Utterance-level Aggregation For Speaker Recognition In The Wild
362 stars 98 forks source link

How to measure the similarity of two utterance ? #53

Closed 553566286 closed 4 years ago

553566286 commented 4 years ago

Hi WeidiXie Thanks for your good job, but there is a question confused me , how did you measure the similarity between two vectors? The Network framework didn't mention this, such like x-vector, the Similarity was measured by PLDA of cosine similarity . But In the predict.py, I saw that you sum (v1*v2)to get the score, what did that mean? regareds

WeidiXie commented 4 years ago

As you pointed out, it's sum(v1 * v2), with v1 and v2 both l2 normalised, meaning, cosine similarity.