Neural Predictive Coding using Convolutional Neural Networks towards Unsupervised Learning of Speaker Characteristics
This addresses the problem of speaker recognition without labeled data, offering a novel unsupervised approach that is incremental but shows competitive results.
The paper tackles unsupervised learning of speaker-specific features from unlabeled audio data, proposing Neural Predictive Coding (NPC) which achieves state-of-the-art performance in short-duration speaker identification and provides complementary information to supervised methods in verification tasks.
Learning speaker-specific features is vital in many applications like speaker recognition, diarization and speech recognition. This paper provides a novel approach, we term Neural Predictive Coding (NPC), to learn speaker-specific characteristics in a completely unsupervised manner from large amounts of unlabeled training data that even contain many non-speech events and multi-speaker audio streams. The NPC framework exploits the proposed short-term active-speaker stationarity hypothesis which assumes two temporally-close short speech segments belong to the same speaker, and thus a common representation that can encode the commonalities of both the segments, should capture the vocal characteristics of that speaker. We train a convolutional deep siamese network to produce "speaker embeddings" by learning to separate `same' vs `different' speaker pairs which are generated from an unlabeled data of audio streams. Two sets of experiments are done in different scenarios to evaluate the strength of NPC embeddings and compare with state-of-the-art in-domain supervised methods. First, two speaker identification experiments with different context lengths are performed in a scenario with comparatively limited within-speaker channel variability. NPC embeddings are found to perform the best at short duration experiment, and they provide complementary information to i-vectors for full utterance experiments. Second, a large scale speaker verification task having a wide range of within-speaker channel variability is adopted as an upper-bound experiment where comparisons are drawn with in-domain supervised methods.