ADAPTIVE: A Novel Dataset For Acoustic DysArthria deTection through temPoral Inference and Voice Engineering

Dysarthria is a prevalent speech disorder affecting approximately 53% of individuals with speech-related challenges, often arising from neurological conditions such as strokes, cerebral palsy, or Parkinson’s disease. This disorder disrupts the coordination and strength of the muscles used for speech, complicating clear communication, especially with unfamiliar listeners. The impact of dysarthria extends beyond mere communication difficulties; it significantly affects social interactions, job prospects, and educational experiences. Consequently, these challenges can diminish the overall quality of life for those affected, making it imperative to address the disorder effectively.

The research questions guiding this study on pre-screening for dysarthria using machine learning techniques are as follows:
RQ1
What specific acoustic features contribute most to detecting speech dysarthria?
RQ2
How can machine learning algorithms be optimized to enhance the accuracy of dysarthria detection compared to traditional assessment methods?
RQ3
Can a minimum number of MFCC Features along with voice engineered features perform equally or better than taking into account a vast number of MFCC Features only?

Hence we introduce ADAPTIVE: A Novel Dataset For Acoustic DysArthria deTection through temPoral Inference and Voice Engineering.

Corresponding paper in the pipeline is yet to be published.

Please cite this dataset of it helps in your studies or if you build your own dataset using the acoustic-temporal feature engineering scripts, ML models or use findings from our research in your papers.

失语症是一种普遍的语言障碍，影响着约53%存在语言相关挑战的个人，通常由中风、脑瘫或帕金森病等神经系统疾病引起。该障碍扰乱了用于言语的肌肉的协调与力量，使得清晰沟通变得复杂，尤其是在面对不熟悉听众时。失语症的影响不仅限于沟通困难；它显著地影响了社交互动、职业前景和教育经历。因此，这些挑战可能会降低受影响者的生活质量，因此，有效应对这一障碍变得至关重要。

本研究的指导性研究问题，旨在利用机器学习技术进行失语症预筛查，具体如下：

RQ1
何种特定的声学特征对检测言语失语症贡献最大？

RQ2
如何优化机器学习算法，以提高失语症检测的准确性，相较于传统评估方法？

RQ3
是否可以使用最小数量的MFCC特征以及语音工程特征，其表现与考虑大量MFCC特征相当或更优？

因此，我们引入了ADAPTIVE：一种基于时间推理和语音工程的新型声学失语症检测数据集。

相关论文尚待发表。

若此数据集有助于您的研究，或您基于声学-时间特征工程脚本、机器学习模型或使用我们的研究发现的论文构建自己的数据集，请引用此数据集。