Chinese Dysarthria Speech Database (CDSD)

Dysarthric speech can severely impact social communication, and accurate dysarthric speech recognition remains challenging largely due to limited data availability. CDSD addresses this gap with 133 hours of recordings from 44 speakers, making it one of the largest Chinese dysarthria datasets to date. Benchmarks report a best Character Error Rate (CER) of 16.4%, highlighting the potential of DSR to improve communication for individuals with dysarthria.