Universal Background Model

A Universal Background Model (UBM) is a large Gaussian Mixture Model trained on speech from many speakers to represent speaker-independent acoustic characteristics. The UBM serves as a reference distribution against which individual speaker models are compared, typically using Maximum A Posteriori (MAP) adaptation. In accessibility contexts, UBMs are foundational to speaker verification systems and iVector extraction techniques used for dysarthric speech assessment, intelligibility prediction, and personalized speech recognition systems that adapt to individual users with speech impairments.