Statistics for An Evaluation of Unsupervised Acoustic Model Training for a Dysarthric Speech Interface