Jun Luo

Jun Luo
research category image

Main area of research
Speech recognition

Selected articles

Discriminative speaker adaptation with eigenvoices

Eigenvoice is an effective speaker adaptation approach and capable of balancing the performance and the requirement for a large amount of adaptation data. However, the

Read more

Efficient keyword spotting system for information retrieval

A new two-stage keyword spotting system was proposed based on syllable graph for audio information retrieval task, which could efficiently spot the interesting words in

Read more

Multigrained model adaptation with MAP and reference speaker weighting for text independent speaker verification

When traditional Maximum a Posteriori (MAP) adaptation is used to adapt a universal background model (UBM), some model components with little or no enrollment data

Read more

Eigenvoice-based MAP adaptation within correlation subspace

In recent years, the eigenvoice approach has proven to be an efficient method for rapid speaker adaptation, which directs the adaptation according to the analysis

Read more

Latent correlation analysis of HMM parameters for speech recognition’

Correlation between HMM parameters has been utilized for various rapid speaker adaptation, e.g. eigenvoice adaptation. The covariance matrix of the supervector which is a concatenation

Read more


1 2