Speaker Verification (Tensorflow-2)

A Tensorflow-2 Repository containing implementation of DNN architectures for Speaker Identification/Speaker Verification. The results of the models along with detailed analysis will be available here. For the purpose of training and testing a smaller version of VoxCeleb dataset containing audio samples of only Indian Celebritites is used in this project in order to facilitate faster prototyping.

Models Implemented

• Baseline CNN Model
• Temporal Convolutional Network
• Siamese Capsule Network(Implementation Here)
• RawNet
• ResDefNet

Dataset

• Indian Celeb Dataset

Paper Links

1. Rawnet: Advanced end-to-end deep neural network using raw waveforms for text-independent speaker verification
2. Siamese Capsule Network For End-to-end Speaker Recognition In The Wild
3. Time-domain Speaker Verification Using Temporal Convolutional Networks
4. Speaker Verification System Based on Deformable CNN and Time-Frequency Attention

Comments

This is a work in progress