Dysarthria

About: Dysarthria is a research topic. Over the lifetime, 2402 publications have been published within this topic receiving 56554 citations.

Towards a Deep Learning Based ASR System for Users with Dysarthria

Abstract: In this paper, we investigate the benefits of deep learning approaches for the development of personalized assistive technology solutions for users with dysarthria, a speech disorder that leads to low intelligibility of users’ speaking. It prevents these people from using automatic speech recognition (ASR) solutions on computers and mobile devices. In order to address these issue, our effort is to leverage convolutional neural networks toward a speaker dependent ASR software solution intended for users with dysarthria, which can be trained according to particular user’s needs and preferences.

The importance of deep speech phenotyping for neurodevelopmental and genetic disorders: a conceptual review

Abstract: Speech is the most common modality through which language is communicated, and delayed, disordered, or absent speech production is a hallmark of many neurodevelopmental and genetic disorders. Yet, speech is not often carefully phenotyped in neurodevelopmental disorders. In this paper, we argue that such deep phenotyping, defined as phenotyping that is specific to speech production and not conflated with language or cognitive ability, is vital if we are to understand how genetic variations affect the brain regions that are associated with spoken language. Speech is distinct from language, though the two are related behaviorally and share neural substrates. We present a brief taxonomy of developmental speech production disorders, with particular emphasis on the motor speech disorders childhood apraxia of speech (a disorder of motor planning) and childhood dysarthria (a set of disorders of motor execution). We review the history of discoveries concerning the KE family, in whom a hereditary form of communication impairment was identified as childhood apraxia of speech and linked to dysfunction in the FOXP2 gene. The story demonstrates how instrumental deep phenotyping of speech production was in this seminal discovery in the genetics of speech and language. There is considerable overlap between the neural substrates associated with speech production and with FOXP2 expression, suggesting that further genes associated with speech dysfunction will also be expressed in similar brain regions. We then show how a biologically accurate computational model of speech production, in combination with detailed information about speech production in children with developmental disorders, can generate testable hypotheses about the nature, genetics, and neurology of speech disorders.Though speech and language are distinct, specific types of developmental speech disorder are associated with far-reaching effects on verbal communication in children with neurodevelopmental disorders. Therefore, detailed speech phenotyping, in collaboration with experts on pediatric speech development and disorders, can lead us to a new generation of discoveries about how speech development is affected in genetic disorders.

Machine Learning in Assistive Technology: a Solution for People with Dysarthria

Abstract: Nowadays, dysarthric speech processing represents a challenge in assistive technology contexts. In this paper, we investigate the use of machine learning in conjunction with convolutional neural networks to implement a speaker dependent solution that is capable to detect just a few number of predefined keywords. The proposed system has been trained with utterances from Italian users with severe and mild dysarthria and it is configurable according to specific users' preferences.

Synthesizing Speech from Intracranial Depth Electrodes using an Encoder-Decoder Framework

Abstract: Speech Neuroprostheses have the potential to enable communication for people with dysarthria or anarthria. Recent advances have demonstrated high-quality text decoding and speech synthesis from electrocorticographic grids placed on the cortical surface. Here, we investigate a less invasive measurement modality in three participants, namely stereotactic EEG (sEEG) that provides sparse sampling from multiple brain regions, including subcortical regions. To evaluate whether sEEG can also be used to synthesize high-quality audio from neural recordings, we employ a recurrent encoder-decoder model based on modern deep learning methods. We find that speech can indeed be reconstructed with correlations up to 0.8 from these minimally invasive recordings, despite limited amounts of training data.

Correlates of older adults’ discrimination of acoustic properties in speech

Abstract: Auditory discrimination of speech stimuli is an essential tool in speech and language therapy, for example, in dysarthria rehabilitation. It is unclear, however, which listener characteristics are associated with the ability to perceive differences between one’s own utterance and target speech. Knowledge about such associations may help to support patients participating in speech and language therapy programs that involve auditory discrimination tasks. Discrimination performance was evaluated in 96 healthy participants over 60 years of age as individuals with dysarthria are typically in this age group. Participants compared meaningful words and sentences on the dimensions of loudness, pitch, and speech rate. Auditory abilities were assessed using pure-tone audiometry, speech audiometry, and speech understanding in noise. Cognitive measures included auditory short-term memory, working memory, and processing speed. Linguistic functioning was assessed by means of vocabulary knowledge and language proficiency. Exploratory factor analyses showed that discrimination performance was primarily associated with cognitive and linguistic skills, rather than auditory abilities. Accordingly, older adults’ discrimination performance was mainly predicted by cognitive and linguistic skills. Discrimination accuracy was higher in older adults with better speech understanding in noise, faster processing speed, and better language proficiency, but accuracy decreased with age. This raises the question whether these associations generalize to clinical populations and, if so, whether patients with better cognitive or linguistic skills may benefit more from discriminationbased therapeutic approaches than patients with poorer cognitive or linguistic abilities.