Subspace distribution clustering hidden Markov model

By E. Bocchieri, Brian Kan-Wing Mak. 2001

117 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Cross-lingual speech recognition under runtime resource constraints (2009, methodology): “The AMs were further compressed using the sub-space coding algorithm [7] to reduce the size and to speed up the decoding.”
Eigentriphones for Context-Dependent Acoustic Modeling (2013, background): “Various HMM parameters have been tied successfully, for example, generalized triphones [9], tied states [10], shared distributions or senones [11], and tied subspace Gaussian distributions [12].”
Small-footprint high-performance deep neural network-based speech recognition using split-VQ (2015, methodology): “For example, given a mini-batch of training feature vectors, Xmb, and the corresponding labels, the weight matrix A(l) is updated using A(l) ← A(l) − ε∂L(X mb) ∂A(l) (6) in which ε is a learning rate.”
Multitask Learning of Deep Neural Networks for Low-Resource Speech Recognition (2015, methodology): “For systems based on hidden Markov modeling (HMM), one common solution is parameter tying or sharing [7], [8], [9], [10], [11].”