Wang Jun, Kothalkar Prasanna V, Kim Myungjong, Bandini Andrea, Cao Beiming, Yunusova Yana, Campbell Thomas F, Heitzman Daragh, Green Jordan R
a Department of Bioengineering , Speech Disorders & Technology Lab.
b Callier Center for Communication Disorders, University of Texas at Dallas , Richardson , TX , USA.
Int J Speech Lang Pathol. 2018 Nov;20(6):669-679. doi: 10.1080/17549507.2018.1508499. Epub 2018 Nov 8.
: This research aimed to automatically predict intelligible speaking rate for individuals with Amyotrophic Lateral Sclerosis (ALS) based on speech acoustic and articulatory samples. Twelve participants with ALS and two normal subjects produced a total of 1831 phrases. NDI Wave system was used to collect tongue and lip movement and acoustic data synchronously. A machine learning algorithm (i.e. support vector machine) was used to predict intelligible speaking rate (speech intelligibility × speaking rate) from acoustic and articulatory features of the recorded samples. Acoustic, lip movement, and tongue movement information separately, yielded a of 0.652, 0.660, and 0.678 and a Root Mean Squared Error (RMSE) of 41.096, 41.166, and 39.855 words per minute (WPM) between the predicted and actual values, respectively. Combining acoustic, lip and tongue information we obtained the highest R (0.712) and the lowest RMSE (37.562 WPM). The results revealed that our proposed analyses predicted the intelligible speaking rate of the participant with reasonably high accuracy by extracting the acoustic and/or articulatory features from one short speech sample. With further development, the analyses may be well-suited for clinical applications that require automatic speech severity prediction.
本研究旨在基于语音声学和发音样本自动预测肌萎缩侧索硬化症（ALS）患者的可理解语速。12名ALS患者和2名正常受试者共生成了1831个短语。使用NDI Wave系统同步收集舌头和嘴唇的运动以及声学数据。使用机器学习算法（即支持向量机）从录制样本的声学和发音特征预测可理解语速（言语可懂度×语速）。单独的声学、嘴唇运动和舌头运动信息，预测值与实际值之间的相关系数分别为0.652、0.660和0.678，均方根误差（RMSE）分别为每分钟41.096、41.166和39.855个单词（WPM）。结合声学、嘴唇和舌头信息，我们获得了最高的相关系数（0.712）和最低的均方根误差（37.562 WPM）。结果表明，我们提出的分析方法通过从一个简短语音样本中提取声学和/或发音特征，以相当高的准确率预测了参与者的可理解语速。随着进一步发展，这些分析可能非常适合需要自动语音严重程度预测的临床应用。
