Speech recognition using the atomic speech units constructed from overlapping articulatory features

By Li Deng, Don X. Sun. 1993

In recent years, the development of a feature‐based general statistical framework has been pursued for automatic speech recognition via novel designs of minimal or atomic units of speech, aiming at a parsimonious scheme to share the interword and interphone speech data and at a unified way to account for the context‐dependent behaviors in speech. The basic design philosophy has been motivated by the theory of distinctive features and by a new form of phonology which argues for use of multidimensional articulatory structures. In this paper, the most recently developed feature‐based recognizer is presented, which is capable of operating on all classes of English sounds. Detailed descriptions of the design considerations for the recognizer and of key aspects of the design process are provided. This process, which is called lexicon ‘‘compilation,’’ consists of three elements (1) establishing a feature‐specification system; (2) constructing a probabilistic and fractional temporal overlapping pattern across the features; and (3) mapping from the feature‐overlap pattern to a state‐transition graph. A standard phonetic classification task from the TIMIT database is used as a test bed to evaluate the performance of the recognizer. The experimental results provide preliminary evidence for the effectiveness of the feature‐based approach to speech recognition.

How researchers see this paper

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

Citations in the corpus

Deep Learning: Methods and Applications (2014, background): “One intermediate neural network layer in the implementation of this detection-based framework explicitly represents the speech attributes, which are simplified entities from the “atomic” units of speech developed in the early work of [101, 355].”
A tutorial survey of architectures, algorithms, and applications for deep learning (2014, background): “One intermediate neural network layer in the implementation of this detection-based framework explicitly represents the speech attributes, which are simplified entities from the “atomic” units of speech developed in the early work of [136, 137].”
A dynamic, feature-based approach to the interface between phonology and phonetics for speech modeling and recognition (1998, methodology): “An earlier version of this model, using piecewise-static trajectories (a conventional HMM) to approximate continuous trajectories in speech acoustics, was described in [28] [30] [29] with several means of evaluation on the TIMIT database .”
Machine Learning Paradigms for Speech Recognition: An Overview (2013, unknown): “, [130], [237], [238].”
Regression-Based Context-Dependent Modeling of Deep Neural Networks for Speech Recognition (2014, methodology): “The proposed work is also different from the detector-based automatic speech recognition approach [22], [23], [24], where NN [22] or DNN [23] with binary outputs are used to detect the articulator attributes in speech and a merger is used to combine various attributes to predict the state output pos”
Modeling Spectral Envelopes Using Restricted Boltzmann Machines and Deep Belief Networks for Statistical Parametric Speech Synthesis (2013, methodology): “The architectures used in deep learning as applied to speech processing have been motivated by the multi-layered structures in both speech production and perception involving phonological features, motor control, articulatory dynamics, and acoustic and auditory parameters [26], [27].”
Switching Dynamic System Models for Speech Articulation and Acoustics (2004, methodology): “We refer the readers to some earlier work for ways of constructing such overlapping units, either by rules or by automatic learning, which have proved e ective in the HMM-like speech recognition framework [20, 18, 17, 51].”
Dynamic Speech Models: Theory, Algorithms, and Applications (2006, methodology): “14 DYNAMIC SPEECH MODELS methods of constructing such overlapping units, either by rules or by automatic learning, can be found in [50, 59, 64–66].”
Boosting attribute and phone estimation accuracies with deep neural networks for detection-based speech recognition (2012, methodology): “Incorporation of asynchrony will significantly modify the attribute targets in running speech in a principled and parsimonious way, as demonstrated in [6][8][9][21].”
Exploiting deep neural networks for detection-based speech recognition (2013, background): “This recognition paradigm is quite flexible and allows the modeling of articulatory asynchrony, yet it suffers from data sparseness problem because the articulatory based HMMs require a large state space.”
An Overview of Deep-Structured Learning for Information Processing (2011, unknown): “One intermediate neural network layer in the implementation of this detection-based framework explicitly represents the speech attributes, which are simplified entities from the ―atomic‖ units of speech developed in the early work of (Deng and Sun, 1994).”
Deep Neural Networks for Acoustic Modeling in Speech Recognition (2012, methodology): “A recent study [65] demonstrated the effectiveness of DBNDNNs for detecting subphonetic speech attributes (also known as phonological or articulatory features [66]) in the widely used Wall Street Journal speech database (5k-WSJ0).”
Computational Models for Speech Production (2018, methodology): “Based mainly on the task-dynamic model originally developed by speech scientists [28] and on our earlier work on overlapping articulatory features [8].”
Context-dependent modelling of deep neural network using logistic regression (2013, methodology): “The proposed work is also different from the detector-based automatic speech recognition approach [14, 15, 16], where NN [14] or DNN [15] with binary outputs are used to detect the articulator attributes in speech and a merger is used to combine various attributes to predict the state output posteri”