# Learning to detect dysarthria from raw speech

arXiv: https://arxiv.org/abs/1811.11101 · HTML: https://mlsanity.com/paper/1811.11101/

## Key facts

- Cited by 7 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 27 Nov 2018.

## Built on by (most cited first, 7 of 7)

- [Machine Learning for Stuttering Identification: Review, Challenges and Future Directions](https://mlsanity.com/paper/2107.04057/) (2021)
- [Comparing Supervised Models And Learned Speech Representations For Classifying Intelligibility Of Disordered Speech On Selected Phrases](https://mlsanity.com/paper/2107.03985/) (2021)
- [Unsupervised Domain Adaptation for Dysarthric Speech Detection via Domain Adversarial Training and Mutual Information Minimization](https://mlsanity.com/paper/2106.10127/) (2021)
- [WavRx: a Disease-Agnostic, Generalizable, and Privacy-Preserving Speech Health Diagnostic Model](https://mlsanity.com/paper/2406.18731/) (2024)
- [Developing Conversational Speech Systems for Robots to Detect Speech Biomarkers of Cognition in People Living with Dementia](https://mlsanity.com/paper/2502.10896/) (2025)
- [Bridging the Perceptual-Statistical Gap in Dysarthria Assessment: Why Machine Learning Still Falls Short](https://mlsanity.com/paper/2510.22237/) (2025)
- [Self-Supervised Speech Representations for Cross-Speaker Dysarthria Detection During Awake Craniotomy](https://mlsanity.com/paper/2610.11825/) (2026)

## Abstract (arXiv)

Speech classifiers of paralinguistic traits traditionally learn from diverse hand-crafted low-level features, by selecting the relevant information for the task at hand. We explore an alternative to this selection, by learning jointly the classifier, and the feature extraction. Recent work on speech recognition has shown improved performance over speech features by learning from the waveform. We extend this approach to paralinguistic classification and propose a neural network that can learn a filterbank, a normalization factor and a compression power from the raw speech, jointly with the rest of the architecture. We apply this model to dysarthria detection from sentence-level audio recordings. Starting from a strong attention-based baseline on which mel-filterbanks outperform standard low-level descriptors, we show that learning the filters or the normalization and compression improves over fixed features by 10% absolute accuracy. We also observe a gain over OpenSmile features by learning jointly the feature extraction, the normalization, and the compression factor with the architecture. This constitutes a first attempt at learning jointly all these operations from raw audio for a speech classification task.