U-vectors: Generating clusterable speaker embedding from unlabeled data
This research addresses the problem of generating speaker embeddings without relying on large pre-trained datasets or domain adaptation, which is beneficial for speaker recognition systems in data-scarce or diverse language environments.
The paper proposes U-vectors, a method to generate clusterable speaker embeddings from unlabeled speech data using an unsupervised training strategy. This approach relies on the assumption that small speech segments contain a single speaker and employs pairwise constraints with noise augmentation to train an AutoEmbedder architecture. The method was evaluated on TIMIT, LibriSpeech, and a Bengali dataset, achieving satisfactory performance.
Speaker recognition deals with recognizing speakers by their speech. Most speaker recognition systems are built upon two stages, the first stage extracts low dimensional correlation embeddings from speech, and the second performs the classification task. The robustness of a speaker recognition system mainly depends on the extraction process of speech embeddings, which are primarily pre-trained on a large-scale dataset. As the embedding systems are pre-trained, the performance of speaker recognition models greatly depends on domain adaptation policy, which may reduce if trained using inadequate data. This paper introduces a speaker recognition strategy dealing with unlabeled data, which generates clusterable embedding vectors from small fixed-size speech frames. The unsupervised training strategy involves an assumption that a small speech segment should include a single speaker. Depending on such a belief, a pairwise constraint is constructed with noise augmentation policies, used to train AutoEmbedder architecture that generates speaker embeddings. Without relying on domain adaption policy, the process unsupervisely produces clusterable speaker embeddings, termed unsupervised vectors (u-vectors). The evaluation is concluded in two popular speaker recognition datasets for English language, TIMIT, and LibriSpeech. Also, a Bengali dataset is included to illustrate the diversity of the domain shifts for speaker recognition systems. Finally, we conclude that the proposed approach achieves satisfactory performance using pairwise architectures.