Initial evaluation of hidden dynamic models on conversational speech

By J. Picone, Sandi Pike, Roland Reagan, T. Kamm, John Scott Bridle, Li Deng, Z. Ma, H. Richards, Mike Schuster. 1999

How researchers see this paper

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

1. Li Deng
2. Heiga Zen
3. Dong Yu
4. Yoshua Bengio
5. Xu Tan
6. Keiichi Tokuda
7. Junichi Yamagishi
8. Geoffrey E. Hinton

Citations in the corpus

• Deep Learning: Methods and Applications (2014, background): “Subsequent work added a new hidden layer into the dynamic model to explicitly account for the target-directed, articulatory-like properties in human speech generation [45, 73, 74, 83, 96, 75, 90, 231, 232, 233, 251, 282].”
• A tutorial survey of architectures, algorithms, and applications for deep learning (2014, background): “In one implementation, lt took the form of shallow MLP weights [69, 163, 164].”
• Continuous latent variable models for dimensionality reduction and sequential data reconstruction (2001, methodology): “In an evaluation of the approach with a conversational speech recognition task with the Switchboard corpus (Picone et al., 1999), improvements in terms of word error rate compared to a standard acoustic HMM only occur if, as well as the most likely hypotheses, the reference transcription is given (w”
• Target-directed mixture dynamic models for spontaneous speech recognition (2004, background): “In recent research work along this direction [4], [8]–[10], [21], we have been developing a statistical coarticulatory model for spontaneous speech recognition.”
• A path-stack algorithm for optimizing dynamic regimes in a statistical hidden dynamic model of speech (2000, unknown): “Details of the model and some earlier experiments are provided in Deng and Ma (1999) and Picone et al. (1999).”
• An Overview of Deep-Structured Learning for Information Processing (2011, background): “Subsequent work added a new hidden layer into the dynamic model so as to explicitly account for the target-directed, articulatory-like properties in human speech generation (Deng and Ramsay, 1997; Deng, 1998; Bridle et al., 1998; Deng, 1999; Picone et al., 1999; Deng, 2003).”
• Machine Learning Paradigms for Speech Recognition: An Overview (2013, background): “There have been several studies on switching nonlinear state space models for ASR, both theoretical [39], [49] and experimental [41]–[43], [50].”
• Dynamic Speech Models: Theory, Algorithms, and Applications (2006, unknown): “, [24, 42, 48, 49, 84, 85, 107–112]); • model with abstract dynamics (e.”
• Reformulating the HMM as a trajectory model by imposing explicit relationships between static and dynamic feature vector sequences (2004, background): “, 1994; Deng and Aksmanovic, 1997; Gish and Ng, 1996), hidden dynamical model (Richards and Bridle, 1999; Picone et al., 1999), vocal tract resonance model (Ma, 2000; Zhou et al.”
• Efficient decoding strategies for conversational speech recognition using a constrained nonlinear state-space model (2003, background): “Digital Object Identifier 10.1109/TSA.2003.818075 reduction, several versions of the global, " super-segmental " speech models described in [10]–[12], [28] have, in various ways, integrated the speaking rate variation with the contextual variation in spontaneous speech.”