Search

Speech is a critical form of human communication and is central to our daily lives. Yet, despite decades of study, an understanding of the fundamental neural control of speech production remains incomplete. Current theories model speech production as a hierarchy from sentences and phrases down to words, syllables, speech sounds (phonemes), and the actions of vocal tract articulators used to produce speech sounds (articulatory gestures). Here, we investigate the cortical representation of articulatory gestures and phonemes in ventral precentral and inferior frontal gyri in men and women. Our results indicate that ventral precentral cortex represents gestures to a greater extent than phonemes, while inferior frontal cortex represents both gestures and phonemes. These findings suggest that speech production shares a common cortical representation with that of other types of movement, such as arm and hand movements. This has important implications both for our understanding of speech production and for the design of brain-machine interfaces to restore communication to people who cannot speak. [ABSTRACT FROM AUTHOR]

Segmental models are an alternative to frame-based models for sequence prediction, where hypothesized path weights are based on entire segment scores rather than a single frame at a time. Neural segmental models are segmental models that use neural network-based weight functions. Neural segmental models have achieved competitive results for speech recognition, and their end-to-end training has been explored in several studies. In this work, we review neural segmental models, which can be viewed as consisting of a neural network-based acoustic encoder and a finite-state transducer decoder. We study end-to-end segmental models with different weight functions, including ones based on frame-level neural classifiers and on segmental recurrent neural networks. We study how reducing the search space size impacts performance under different weight functions. We also compare several loss functions for end-to-end training. Finally, we explore training approaches, including multistage versus end-to-end training and multitask training that combines segmental and frame-level losses. [ABSTRACT FROM PUBLISHER]

Aspects of speech production have provided inspiration for ideas in speech technologies throughout the history of speech processing research. This special issue was inspired by the 2013Workshop on Speech Production in Automatic Speech Recognition in Lyon, France, and this introduction provides an overview of the included papers in the context of the current research landscape. [ABSTRACT FROM AUTHOR]

Modern automatic speech recognition systems handle large vocabularies of words, making it infeasible to collect enough repetitions of each word to train individual word models. Instead, large-vocabulary recognizers represent each word in terms of subword units. Typically the subword unit is the phone, a basic speech sound such as a single consonant or vowel. Each word is then represented as a sequence, or several alternative sequences, of phones specified in a pronunciation dictionary. Other choices of subword units have been studied as well. The choice of subword units, and the way in which the recognizer represents words in terms of combinations of those units, is the problem of subword modeling. Different subword models may be preferable in different settings, such as high-variability conversational speech, highnoise conditions, low-resource settings, or multilingual speech recognition. This article reviews past, present, and emerging approaches to subword modeling. To make clean comparisons between many approaches, the review uses the unifying language of graphical models. [ABSTRACT FROM PUBLISHER]

Although much is known about how speech is produced, and research into speech production has resulted in measured articulatory data, feature systems of different kinds, and numerous models, speech production knowledge is almost totally ignored in current mainstream approaches to automatic speech recognition. Representations of speech production allow simple explanations for many phenomena observed in speech which cannot be easily analyzed from either acoustic signal or phonetic transcription alone. In this article, a survey of a growing body of work in which such representations are used to improve automatic speech recognition is provided. [ABSTRACT FROM AUTHOR]

Three research prototype speech recognition systems are described, all of which use recently developed methods from artificial intelligence (specifically support vector machines, dynamic Bayesian networks, and maximum entropy classification) in order to implement, in the form of an automatic speech recognizer, current theories of human speech perception and phonology (specifically landmark-based speech perception, nonlinear phonology, and articulatory phonology). All three systems begin with a high-dimensional multiframe acoustic-to-distinctive feature transformation, implemented using support vector machines trained to detect and classify acoustic phonetic landmarks. Distinctive feature probabilities estimated by the support vector machines are then integrated using one of three pronunciation models: a dynamic programming algorithm that assumes canonical pronunciation of each word, a dynamic Bayesian network implementation of articulatory phonology, or a discriminative pronunciation model trained using the methods of maximum entropy classification. Log probability scores computed by these models are then combined, using log-linear combination, with other word scores available in the lattice output of a first-pass recognizer, and the resulting combination score is used to compute a second-pass speech recognition output.