TristouNet: Triplet Loss for Speaker Turn Embedding
This addresses speaker comparison and change detection in speech processing, with incremental improvements over existing methods.
The paper tackled the problem of comparing speech sequences for speaker comparison and speaker change detection by introducing TristouNet, a neural network using triplet loss to project speech into a Euclidean space, resulting in significant improvements over state-of-the-art techniques on short speech turn tasks.
TristouNet is a neural network architecture based on Long Short-Term Memory recurrent networks, meant to project speech sequences into a fixed-dimensional euclidean space. Thanks to the triplet loss paradigm used for training, the resulting sequence embeddings can be compared directly with the euclidean distance, for speaker comparison purposes. Experiments on short (between 500ms and 5s) speech turn comparison and speaker change detection show that TristouNet brings significant improvements over the current state-of-the-art techniques for both tasks.