A speech-controlled environmental control system for people with severe dysarthria

Area of Science:

• * Assistive Technology
• * Speech Recognition
• * Biomedical Engineering

Background:

• * Off-the-shelf automatic speech recognition (ASR) systems perform poorly for individuals with severe dysarthria due to speech variability.
• * Existing systems lack sufficient tolerance for the articulation inconsistencies common in dysarthric speech.
• * There is a need for specialized ASR solutions to improve control of electronic assistive technology for this population.

Purpose of the Study:

• * To develop and evaluate a speaker-dependent, limited-vocabulary ASR application with enhanced tolerance for speech variability.
• * To assess a computerized training package designed to improve vocalization consistency in dysarthric speakers and augment recognizer training data.
• * To implement and test a speech-controlled environmental control system (ECS) using the developed ASR technology.

Main Methods:

• * Development of a specialized ASR application with a limited vocabulary and speaker-dependent training.
• * Implementation of a computerized training program to enhance speech consistency and provide additional training data.
• * Field trials of the speech-controlled ECS with individuals experiencing severe dysarthria.

Main Results:

• * User training significantly improved speech recognition rates from 88.5% to 95.4% (p<0.001).
• * The system achieved good recognition rates (mean 86.9%) for individuals with severe dysarthria in home environments.
• * Speech-controlled ECS demonstrated faster task completion times (7.7s vs. 16.9s, p<0.001) compared to switch-scanning systems, despite lower task completion accuracy (78.6% vs. 94.8%).

Conclusions:

• * The developed speech-controlled environmental control system is a viable alternative to switch-scanning systems for some individuals with severe dysarthria.
• * This ASR-based system can lead to more efficient control of the home environment for users with severe dysarthria.
• * The combination of a tolerant ASR system and a targeted training program effectively addresses the challenges of speech control for individuals with dysarthria.