Comparing Human and Automatic Recognition of Dutch Dysarthric Continuous Speech: A Case Study

2026-06-29

AI Summary

This study addresses the significant challenge posed by Dutch continuous speech from a speaker with severe dysarthria, which yields extremely high word error rates (WER) for both human listeners and existing automatic speech recognition (ASR) systems. For the first time, it systematically compares human performance against three state-of-the-art ASR systems—Whisper-large-V3, Google Chirp 3, and Omnilingual—on both read and spontaneous speech from this speaker, and further develops personalized ASR models through speaker-specific fine-tuning. The fine-tuned models achieve WERs below 23%, substantially outperforming human listeners, whose WER exceeds 70%. These results not only confirm the efficacy of individualized fine-tuning but also demonstrate, for the first time in Dutch severe dysarthric speech, ASR performance surpassing human comprehension, marking a critical advance toward practical assistive communication technologies.

Abstract

In our goal to develop personalised dysarthric speech recognition (DSR) models, this study compared the recognition performances of human listeners and those of three state-of-the-art, off-the-shelf ASR systems (Whisper-large-V3, Google Chirp 3, and Omnilingual) on the recognition of Dutch continuous read and spontaneous speech from a single speaker with severe dysarthria. Results showed that both humans listeners and the three off-the-shelf ASR systems exhibit word error rates (WER) exceeding 70% on average, indicating that DSR is highly challenging for both humans and ASR systems. Fine-tuning on the dysarthric speech significantly reduced WER. Although overall WERs are still quite high (>23%), the personalised DSR models outperformed the human listeners, and performance is getting closer to being useful for supporting day-to-day communication of dysarthric speakers. Future research should focus on improving personalized DSR on spontaneous speech and longer utterances in the case of read speech, with a specific focus on particular phonemes.

Problem

• dysarthric speech recognition
• continuous speech
• word error rate
• personalized ASR
• Dutch

Innovation

• personalized dysarthric speech recognition
• fine-tuning
• word error rate
• automatic speech recognition
• dysarthria