Deep Speaker Feature Learning for Text-independent Speaker Verification
This work addresses speaker verification for security or biometric applications, but it is incremental as it builds on existing deep learning approaches with a novel network structure.
The paper tackled the problem of learning high-quality speaker features for text-independent speaker verification by proposing a convolutional time-delay deep neural network (CT-DNN). The result showed that even with a single 0.3-second feature, the equal error rate (EER) could be as low as 7.68% on the Fisher database.
Recently deep neural networks (DNNs) have been used to learn speaker features. However, the quality of the learned features is not sufficiently good, so a complex back-end model, either neural or probabilistic, has to be used to address the residual uncertainty when applied to speaker verification, just as with raw features. This paper presents a convolutional time-delay deep neural network structure (CT-DNN) for speaker feature learning. Our experimental results on the Fisher database demonstrated that this CT-DNN can produce high-quality speaker features: even with a single feature (0.3 seconds including the context), the EER can be as low as 7.68%. This effectively confirmed that the speaker trait is largely a deterministic short-time property rather than a long-time distributional pattern, and therefore can be extracted from just dozens of frames.