Robust Feature Clustering for Unsupervised Speech Activity Detection
This addresses the need for speech activity detection in zero-resource or low-resource scenarios where annotations are unavailable, though it is incremental as it builds on existing clustering and statistical methods.
The paper tackles the problem of speech activity detection without annotated data by proposing a fully unsupervised clustering approach using the Hartigan dip test, achieving superior performance over a two-component GMM baseline on NIST OpenSAD 2015 and NIST OpenSAT 2017 datasets.
In certain applications such as zero-resource speech processing or very-low resource speech-language systems, it might not be feasible to collect speech activity detection (SAD) annotations. However, the state-of-the-art supervised SAD techniques based on neural networks or other machine learning methods require annotated training data matched to the target domain. This paper establish a clustering approach for fully unsupervised SAD useful for cases where SAD annotations are not available. The proposed approach leverages Hartigan dip test in a recursive strategy for segmenting the feature space into prominent modes. Statistical dip is invariant to distortions that lends robustness to the proposed method. We evaluate the method on NIST OpenSAD 2015 and NIST OpenSAT 2017 public safety communications data. The results showed the superiority of proposed approach over the two-component GMM baseline. Index Terms: Clustering, Hartigan dip test, NIST OpenSAD, NIST OpenSAT, speech activity detection, zero-resource speech processing, unsupervised learning.