If you are interested in using MDSA dataset, please take the following steps:
1. All current publicly available MDSA datasets are listed below with descriptions. You may select the ones you wish to download by check the checkboxes on the left. Then click the “Next” button.
2. On the next page, you will have to download, print, sign and scan an EULA (End User License Agreement) and upload it through the link provided. Then click the "Submit" button.
3. Upon receipt of the EULA, the access of selected dataset files will be granted by dataset administrators. Depend on the workload of the dataset administrators, this could take a couple of days.
4. You may log in your personal page to check whether the accesses are granted. Once you do, you may download the requested files through your MDSA website "HowToAccess" section.
The main dataset
·The main MDSA dataset contains 17.05 hours of synchronized audio and video data, obtained from 89 native speakers of Mandarin Chinese.
·64 subacute stroke participants and 25 normal controls.
·all audio-visual data were manually annotated and simultaneously verified.
·include different levels of speech materials, ranging from syllables, characters, words, sentences, to spontaneous speech
·comprehensive clinical evaluations of each patient have also been provided, such as FDA assessment and MoCA
File size: Train(2.9G) and Dev(217MB)
Latest publication based on this dataset:
·Liu, J., Du, X., Lu, S., Zhang, Y. M., An-ming, H. U., Ng, M. L., ... & Yan, N. (2023). Audio-video database from subacute stroke patients for dysarthric speech intelligence assessment and preliminary analysis. Biomedical Signal Processing and Control, 79, 104161.
·Liu X, Du X, Liu J, et al. Automatic Assessment of Dysarthria Using Audio-visual Vowel Graph Attention Network[J]. arXiv preprint arXiv:2405.03254, 2024.
·Lu, S., Du, X., Liu, J., Zhang, Y. M., Zhao, S., Su, R., ... & Yan, N. (2022, October). A New Method for Predicting Severity Level of Dysarthric Speech Based on Joint Feature-Sample Selection using Audio-Visual Data. In 2022 International Conference on Asian Language Processing (IALP) (pp. 190-195). IEEE
🔥🔥🔥Access to Dataset【Import】:
[First Step]: Download user agreement: Download Agreement
[Second Step]: Please submit the agreement to your email : huanraozhineng2@siat.ac.cn
[Third Step]: Get the account password from email and click "Next" Button