• Publications
  • Influence
Discriminant Waveletfaces and Nearest Feature Classifiers for Face Recognition
TLDR
Feature extraction, discriminant analysis, and classification rules are three crucial issues for face recognition. Expand
  • 483
  • 53
  • PDF
Adaptive Bayesian Latent Semantic Analysis
TLDR
In this paper, a novel Bayesian PLSA framework is presented. Expand
  • 84
  • 8
Latent Dirichlet learning for document summarization
TLDR
We present a new hierarchical representation of words, sentences and documents in a corpus, and infers the Dirichlet distributions for latent topics and latent themes in word level and sentence level, respectively. Expand
  • 57
  • 5
  • PDF
Dirichlet Class Language Models for Speech Recognition
TLDR
This work presents a new Dirichlet class language model (DCLM), which projects the sequence of history words onto a latent class space and calculates a marginal likelihood over the uncertainties of classes, which are expressed by Dirichlets. Expand
  • 63
  • 4
Convex Divergence ICA for Blind Source Separation
TLDR
We present a novel convex divergence measure for unsupervised learning of independent components and apply it for blind source separation. Expand
  • 32
  • 4
  • PDF
A new eigenvoice approach to speaker adaptation
TLDR
We present two approaches to improve the eigenvoice-based speaker adaptation. Expand
  • 10
  • 4
  • PDF
Large-Vocabulary Continuous Speech Recognition Systems: A Look at Some Recent Advances
TLDR
Large vocabulary continuous speech recognition is far from being solved: background noise, channel distortions, foreign accents, casual and disfluent speech, or unexpected topic change can cause automated systems to make egregious recognition errors. Expand
  • 102
  • 3
  • PDF
A new independent component analysis for speech recognition and separation
TLDR
This paper presents a novel nonparametric likelihood ratio (NLR) objective function for independent component analysis (ICA). Expand
  • 59
  • 3
  • PDF
Mixture of PLDA for Noise Robust I-Vector Speaker Verification
TLDR
In real-world environments, noisy utterances with variable noise levels are recorded and then converted to i-vectors for cosine distance or PLDA scoring. Expand
  • 45
  • 3
  • PDF
Predictive hidden Markov model selection for speech recognition
TLDR
This paper surveys a series of model selection approaches and presents a novel predictive information criterion (PIC) for hidden Markov model (HMM) selection. Expand
  • 42
  • 3