Automated Forced-Choice Tests of Speech Recognition

Area of Science:

• Audiology
• Speech-language pathology
• Biomedical engineering

Background:

• Traditional speech recognition tests require manual scoring and audiologist supervision.
• Developing automated methods can improve efficiency and accessibility in audiological assessments.

Purpose of the Study:

• To create automated tests for speech-recognition threshold (SRT) and word recognition using computerized scoring.
• To ensure automated SRT scores align with pure-tone audiometry averages.
• To develop an automated word recognition test that differentiates normal hearing from sensorineural hearing loss for hearing aid evaluations.

Main Methods:

• An automated SRT protocol was designed to find the lowest response level for spondees.
• A closed-set, touchscreen word recognition test with rhyming and nonrhyming foils was administered.
• Participants included young adults with normal hearing and individuals with sensorineural hearing loss.

Main Results:

• Automated SRT scores closely matched pure-tone averages, similar to conventional methods.
• Word recognition scores were highest with closed-set, nonrhyming foils and varied widely in hearing-impaired participants.
• A moderate correlation was found between word recognition scores and pure-tone thresholds, particularly with higher frequencies.

Conclusions:

• Automated SRTs are comparable to conventional clinical methods.
• The automated closed-set word recognition test effectively distinguishes between normal and hearing-impaired groups.
• These automated procedures offer reliable clinical test results independent of audiologist availability.