AIVIS
Artificial Intelligence
Voice Stethoscope
Phonetic Transcription Model for Korean Clinical Speech Assessment
AIVIS is capable of accurately transcribing Korean phonological variations that are difficult for conventional ASR (Automatic Speech Recognition) systems to process, including phoneme substitution, deletion, tensification, and consonant cluster simplification. Beyond standard pronunciation, AIVIS records speech exactly as it is perceived auditorily, capturing regional dialects, incomplete utterances, and non-standard phonological patterns. This enables the generation of highly precise speech data that can be directly utilized for clinical analysis. These capabilities are powered by a specialized transcription model trained on extensive datasets of Korean standard language, regional dialects, and clinical speech data.
Speech Signal-Based Clinical Assessment System
AIVIS goes beyond transcription by directly analyzing the speech signal itself to derive a wide range of clinical metrics, including pronunciation accuracy—such as Percentage of Consonants Correct (PCC) and Percentage of Vowels Correct (PVC)—as well as resonance, prosody, and overall voice quality. In addition, AIVIS quantifies speech motor task–based measures such as Maximum Phonation Time (MPT), Alternating Motion Rate (AMR), and Sequential Motion Rate (SMR), enabling the detection of voice-based indicators related to functional decline and fatigue of the speech production mechanism. These metrics allow for the extraction of potential clinical biomarkers associated with neurological conditions such as dementia and normal pressure hydrocephalus (NPH). Through this approach, AIVIS provides a critical foundation for early diagnosis and treatment-effect monitoring in patients with brain injury, dysarthria, and related speech disorders.
Multi-Context Application Platform
IVIS is designed to be applicable across diverse populations and settings, ranging from cognitive function assessment in older adults to language development monitoring in children. For older adults, AIVIS performs fine-grained analysis of subtle articulation errors and phonatory impairments that may arise from conditions such as dementia, normal pressure hydrocephalus, and cerebrovascular diseases. This supports early screening and informed treatment planning. For children, AIVIS transcribes immature or developing speech as it is produced, enabling objective assessment of language development levels and seamless integration with personalized articulation training content. Furthermore, AIVIS supports multiple deployment formats—including web, mobile, and clinical environments—allowing for stable and scalable implementation across clinical, home, and research settings.