Voice recognition: an enabling technology for modern health care?

Recent performance breakthroughs in affordable, large vocabulary, speaker independent voice recognition systems have rekindled widespread interest in using voice recognition technology to enhance the palatability and effectiveness of clinician-mediated computing. However, even if industry fully addresses the formidable hardware requirements, less than perfect recognition accuracies, discrete voice recognition requirements, and throughput limitations, there are significant cognitive and implementation issues that must be adequately resolved before voice can become a ubiquitous input modality. Cognitive issues include making allowances for individual differences in verbal communication style and skill levels, the relative cognitive load of using a voice enabled interface compared to alternative modalities, and the user's cognitive style. Implementation issues include a significant training requirement, limited portability, lengthy user switching time, questionable privacy, satisfying hardware requirements and the suitability of voice recognition in specific work environments. The inevitable resolution of these issues coupled with continuously improving voice recognition performance, promises a new era for voice recognition in medicine.