An Automated Lexical Stress Classification Tool for Assessing Dysprosody in Childhood Apraxia of Speech

Childhood apraxia of speech (CAS) commonly affects the production of lexical stress contrast in polysyllabic words. Automated classification tools have the potential to increase reliability and efficiency in measuring lexical stress. Here, factors affecting the accuracy of a custom-built deep neural network (DNN)-based classification tool are evaluated. Sixteen children with typical development (TD) and 26 with CAS produced 50 polysyllabic words. Words with strong–weak (SW, e.g., dinosaur) or WS (e.g., banana) stress were fed to the classification tool, and the accuracy measured (a) against expert judgment, (b) for speaker group, and (c) with/without prior knowledge of phonemic errors in the sample. The influence of segmental features and participant factors on tool accuracy was analysed. Linear mixed modelling showed significant interaction between group and stress type, surviving adjustment for age and CAS severity. For TD, agreement for SW and WS words was >80%, but CAS speech was higher for SW (>80%) than WS (~60%). Prior knowledge of segmental errors conferred no clear advantage. Automatic lexical stress classification shows promise for identifying errors in children’s speech at diagnosis or with treatment-related change, but accuracy for WS words in apraxic speech needs improvement. Further training of algorithms using larger sets of labelled data containing impaired speech and WS words may increase accuracy.

Optimising Speaker-Dependent Feature Extraction Parameters to Improve Automatic Speech Recognition Performance for People with Dysarthria

Within the field of Automatic Speech Recognition (ASR) systems, facing impaired speech is a big challenge because standard approaches are ineffective in the presence of dysarthria. The first aim of our work is to confirm the effectiveness of a new speech analysis technique for speakers with dysarthria. This new approach exploits the fine-tuning of the size and shift parameters of the spectral analysis window used to compute the initial short-time Fourier transform, to improve the performance of a speaker-dependent ASR system. The second aim is to define if there exists a correlation among the speaker’s voice features and the optimal window and shift parameters that minimises the error of an ASR system, for that specific speaker. For our experiments, we used both impaired and unimpaired Italian speech. Specifically, we used 30 speakers with dysarthria from the IDEA database and 10 professional speakers from the CLIPS database. Both databases are freely available. The results confirm that, if a standard ASR system performs poorly with a speaker with dysarthria, it can be improved by using the new speech analysis. Otherwise, the new approach is ineffective in cases of unimpaired and low impaired speech. Furthermore, there exists a correlation between some speaker’s voice features and their optimal parameters.

Can Haptic Stimulation Enhance Music Perception in Hearing-Impaired Listeners?

Cochlear implants (CIs) have been remarkably successful at restoring hearing in severely-to-profoundly hearing-impaired individuals. However, users often struggle to deconstruct complex auditory scenes with multiple simultaneous sounds, which can result in reduced music enjoyment and impaired speech understanding in background noise. Hearing aid users often have similar issues, though these are typically less acute. Several recent studies have shown that haptic stimulation can enhance CI listening by giving access to sound features that are poorly transmitted through the electrical CI signal. This “electro-haptic stimulation” improves melody recognition and pitch discrimination, as well as speech-in-noise performance and sound localization. The success of this approach suggests it could also enhance auditory perception in hearing-aid users and other hearing-impaired listeners. This review focuses on the use of haptic stimulation to enhance music perception in hearing-impaired listeners. Music is prevalent throughout everyday life, being critical to media such as film and video games, and often being central to events such as weddings and funerals. It represents the biggest challenge for signal processing, as it is typically an extremely complex acoustic signal, containing multiple simultaneous harmonic and inharmonic sounds. Signal-processing approaches developed for enhancing music perception could therefore have significant utility for other key issues faced by hearing-impaired listeners, such as understanding speech in noisy environments. This review first discusses the limits of music perception in hearing-impaired listeners and the limits of the tactile system. It then discusses the evidence around integration of audio and haptic stimulation in the brain. Next, the features, suitability, and success of current haptic devices for enhancing music perception are reviewed, as well as the signal-processing approaches that could be deployed in future haptic devices. Finally, the cutting-edge technologies that could be exploited for enhancing music perception with haptics are discussed. These include the latest micro motor and driver technology, low-power wireless technology, machine learning, big data, and cloud computing. New approaches for enhancing music perception in hearing-impaired listeners could substantially improve quality of life. Furthermore, effective haptic techniques for providing complex sound information could offer a non-invasive, affordable means for enhancing listening more broadly in hearing-impaired individuals.

Syntactic aspects of impaired speech in interface with prosody and information structure

This paper aims at studying how the syntactic component of language develops in the speech of people with Broca’s and Wernicke’s aphasia in interface with prosody and informational structure. The data consists of two short interviews in English with aphasic patients. Broca’s aphasia is characterized by the difficulty in processing and producing syntactic structures. In Wernicke’s aphasia, the semantic component is affected, which ends up generating a disconnected and meaningless speech. It was found that in Broca’s aphasia the patient marked some heads of English in final position – as head-final similar to languages like Japanese – instead of head-first, a common parameter of English. In Wernicke’s aphasia, there were some inadequacies in the use of adjuncts and complements that resulted in semantic anomalies.

Automatic speech recognition in neurodegenerative disease

Automatic speech recognition (ASR) could potentially improve communication by providing transcriptions of speech in real time. ASR is particularly useful for people with progressive disorders that lead to reduced speech intelligibility or difficulties performing motor tasks. ASR services are usually trained on healthy speech and may not be optimized for impaired speech, creating a barrier for accessing augmented assistance devices. We tested the performance of three state-of-the-art ASR platforms on two groups of people with neurodegenerative disease and healthy controls. We further examined individual differences that may explain errors in ASR services within groups, such as age and sex. Speakers were recorded while reading a standard text. Speech was elicited from individuals with multiple sclerosis, Friedreich’s ataxia, and healthy controls. Recordings were manually transcribed and compared to ASR transcriptions using Amazon Web Services, Google Cloud, and IBM Watson. Accuracy was measured as the proportion of words that were correctly classified. ASR accuracy was higher for controls than clinical groups, and higher for multiple sclerosis compared to Friedreich’s ataxia for all ASR services. Amazon Web Services and Google Cloud yielded higher accuracy than IBM Watson. ASR accuracy decreased with increased disease duration. Age and sex did not significantly affect ASR accuracy. ASR faces challenges for people with neuromuscular disorders. Until improvements are made in recognizing less intelligible speech, the true value of ASR for people requiring augmented assistance devices and alternative communication remains unrealized. We suggest potential methods to improve ASR for those with impaired speech.

Ataxic syndrome in a chronic phenytoin user

Introduction: Phenytoin is an anticonvulsant used routinely for about eight decades. However, depending on the dose and plasma concentration, its use may be associated with side effects due to toxicity, such as ataxic syndrome. Case report: We attended a 37-year-old patient, epileptic since childhood, who had been using Divalproate Sodium 250mg 8 / 8h, Phenobarbital 150mg once a day and Phenytoin 100mg 8/8 for a long time. He denied smoking and drinking. He was admitted due to acute rotational vertigo, nausea, motor incoordination and impaired speech and gait, progressing for 7 days. Neurological examination revealed drunken dysarthria, pendular patellar reflexes, signs of axial and appendicular incoordination and ataxic gait. Cranial nerves: there was decomposition of the eye movement and hypometric saccades to the vertical upward look, horizontal nystagmus with alternating phases to the extreme looks. Laboratory exams, cranial tomography with and without contrast, brain magnetic resonance with gadolinium): within the normal range. Given the above, our diagnostic hypothesis was Phenytoin Poisoning. After gradually replacing it, there was a progressive improvement in the neurological condition. His serum level was not determined due to the unavailability of this test in our service. Conclusion: In view of its routine use, excluding other etiologies, this diagnosis should always be remembered.

Speech-to-Speech Conversion

A novel approach to build a speech-to-speech conversion (STSC) system for individuals with speech impairment dysarthria is described. STSC system takes impaired speech having inherent disturbance as input and produces a synthesized output speech with good pronunciation and noise free utterance. The STSC system involves two stages, namely automatic speech recognition (ASR) and automatic speech synthesis. ASR transforms speech into text, while automatic speech synthesis (or text-to-speech [TTS]) performs the reverse task. At present, the recognition system is developed for a small vocabulary of 50 words and the accuracy of 94% is achieved for normal speakers and 88% for speakers with dysarthria. The output speech of TTS system has achieved a MOS value of 4.5 out of 5 as obtained by averaging the response of 20 listeners. This method of STSC would be an augmentative and alternative communication aid for speakers with dysarthria.

Inaccurate cortical tracking of speech in adults with impaired speech perception in noise

Impaired speech perception in noise despite normal peripheral auditory function is a common problem in young adults. Despite a growing body of research, the pathophysiology of this impairment remains unknown. This magnetoencephalography study characterizes the cortical tracking of speech in a multi-talker background in a group of highly selected adult subjects with impaired speech perception in noise without peripheral auditory dysfunction. Magnetoencephalographic signals were recorded from 13 subjects with impaired speech perception in noise (six females, mean age: 30 years) and matched healthy subjects while they were listening to 5 different recordings of stories merged with a multi-talker background at different signal to noise ratios (No Noise, +10, +5, 0 and −5 dB). The cortical tracking of speech was quantified with coherence between magnetoencephalographic signals and the temporal envelope of (i) the global auditory scene (i.e. the attended speech stream and the multi-talker background noise), (ii) the attended speech stream only and (iii) the multi-talker background noise. Functional connectivity was then estimated between brain areas showing altered cortical tracking of speech in noise in subjects with impaired speech perception in noise and the rest of the brain. All participants demonstrated a selective cortical representation of the attended speech stream in noisy conditions, but subjects with impaired speech perception in noise displayed reduced cortical tracking of speech at the syllable rate (i.e. 4–8 Hz) in all noisy conditions. Increased functional connectivity was observed in subjects with impaired speech perception in noise in Noiseless and speech in noise conditions between supratemporal auditory cortices and left-dominant brain areas involved in semantic and attention processes. The difficulty to understand speech in a multi-talker background in subjects with impaired speech perception in noise appears to be related to an inaccurate auditory cortex tracking of speech at the syllable rate. The increased functional connectivity between supratemporal auditory cortices and language/attention-related neocortical areas probably aims at supporting speech perception and subsequent recognition in adverse auditory scenes. Overall, this study argues for a central origin of impaired speech perception in noise in the absence of any peripheral auditory dysfunction.