Analysis of MLP-Based Hierarchical Phoneme Posterior Probability Estimator

By Joel Pinto, Garimella S. V. S. Sivaram, Mathew Magimai-Doss, Hynek Hermansky, Hervé Bourlard. 2011

86 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Signal Processing

How researchers see this paper

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

In Artificial Intelligence

1. Li Deng
2. László Tóth
3. Xiao Li
4. Dong Yu
5. Kris Demuynck
6. Fabian Triefenbach
7. Ming Li
8. Mathew Magimai-Doss

In Signal Processing

1. Li Deng
2. László Tóth
3. Xiao Li
4. Dong Yu
5. Ming Li
6. Mathew Magimai-Doss
7. Philip N. Garner
8. Kris Demuynck

Citations in the corpus

• An Overview of Deep-Structured Learning for Information Processing (2011, background): “…of deep neural network models forms the basis of the popular ―tandem‖ approach, where a discriminatively learned neural network is developed in the context of computing discriminant emission probabilities for HMMs; for the most recent work, see (Pinto et al., 2011; Ketabdar and Bourlard, 2010).”
• Current trends in multilingual speech processing (2011, methodology): “In Pinto et al (2011) and Pinto (2010), we have studied the role of the second MLP layer in such hierarchical arrangements using Volterra series and have found that it is predominantly responsible for learning phonetic-temporal patterns present in the posterior features.”
• Convolutional deep maxout networks for phone recognition (2014, methodology): “This architecture was motivated by the success of hierarchical ANN models, where a neural network is trained on a block of acoustic feature vectors, and then a second network is trained on a block of posterior output vectors got from the first network [15, 16].”
• Machine Learning Paradigms for Speech Recognition: An Overview (2013, methodology): “The use of this type of discriminative model for ASR has been documented and summarized in detail in [90]–[92] and analyzed recently in [93].”
• Acoustic Modeling With Hierarchical Reservoirs (2013, background): “Instead of creating deep networks with unsupervised pre-training, [17] and [18] propose a hierarchical MLP consisting of only two MLPs which are both trained in a supervised way, again one after the other.”
• Speaker verification and spoken language identification using a generalized i-vector framework with phonetic tokenizations and tandem features (2014, methodology): “This is also motivated by the hierarchical phoneme posterior probability estimator in [15].”