Abstract:
Voice recognition software is considered a viable option for adaptive computer access for people with disabilities. People with dysarthria, however, often have difficulty using this technology. This study identified specific voice qualities and analyzed their effect on performance of voice recognition software. Results indicated a strong correlation between intelligibility of dysarthric speech and successful use of DragonDictate voice recognition software. The correlation held for three variables within the study: AIDS Word Test, AIDS Sentence Test, and Kent Word List. The strongest correlation was seen using the Kent Word List. A correlation between level of education and successful use of voice recognition was also found, with a higher education level corresponding to successful use. This correlation, while statistically significant, was not as strong as the correlation of intelligibility. Data analysis of characteristics of speech deficits (errors of place, manner, and voice) were not statistically significant. Subjects tended to exhibit a wide variety of speech characteristics, making the significance of a small number of deficits impossible to tease out. However, the strong correlation of intelligibility provides us with a direct measure of potential use for voice recognition independent of specific speech characteristics and potentially easier to measure.