English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Autism spectrum disorder (ASD) is a neurodevelopmental disorder which resultsin altered behavior, social development, and communication patterns. In pastyears, autism prevalence has tripled, with 1 in 54 children now affected. Giventhat traditional diagnosis is a lengthy, labor-intensive process, significantattention has been given to developing systems that automatically screen forautism. Prosody abnormalities are among the clearest signs of autism, withaffected children displaying speech idiosyncrasies including echolalia,monotonous intonation, atypical pitch, and irregular linguistic stresspatterns. In this work, we present a suite of machine learning approaches todetect autism in self-recorded speech audio captured from autistic andneurotypical (NT) children in home environments. We consider three methods todetect autism in child speech: first, Random Forests trained on extracted audiofeatures (including Mel-frequency cepstral coefficients); second, convolutionalneural networks (CNNs) trained on spectrograms; and third, fine-tuned wav2vec2.0--a state-of-the-art Transformer-based ASR model. We train our classifierson our novel dataset of cellphone-recorded child speech audio curated fromStanford's Guess What? mobile game, an app designed to crowdsource videos ofautistic and neurotypical children in a natural home environment. The RandomForest classifier achieves 70% accuracy, the fine-tuned wav2vec 2.0 modelachieves 77% accuracy, and the CNN achieves 79% accuracy when classifyingchildren's audio as either ASD or NT. Our models were able to predict autismstatus when training on a varied selection of home audio clips withinconsistent recording quality, which may be more generalizable to real worldconditions. These results demonstrate that machine learning methods offerpromise in detecting autism automatically from speech without specializedequipment.

Classifying Autism from Crowdsourced Semi-Structured Speech Recordings: A Machine Learning Approach