Naver Cloud Debuts Voice AI to Detect Dementia Symptoms
Naver Cloud researchers presented two papers at the international speech conference Interspeech 2026 in Australia, demonstrating an artificial intelligence system that analyzes speaking speed and pauses to distinguish dementia patients from healthy individuals. The technology examines both the content and manner of speech to identify potential cognitive health issues through phone conversations.
Non-verbal speech features improve diagnostic performance
Traditional speech recognition technology typically filters out pauses, hesitations, and word-recall gaps as background noise. However, Naver Cloud’s new approach treats these non-verbal features as critical diagnostic clues. The research team extracted speech features using OpenAI’s Whisper model and analyzed vocabulary, sentence structure, and narrative coherence using the large language model GPT-5.2. According to the company’s presentation in Australia, diagnostic performance increased from 81.48 percent using transcribed text alone to 90.14 percent after sequentially incorporating narrative themes, speech speed, pauses, and pronunciation data.
Naver Cloud tests technology in clinical trials
Naver Cloud is currently conducting clinical trials with Emocog and Seoul Boramae Hospital to evaluate whether the technology can reliably detect early cognitive decline through phone conversations alone. The company previously showcased its ‘Naver Care Call’ system at the Local Government and Public Week 2026 held in Tokyo, Japan, in May. Meanwhile, market research firm The Business Research Company projected the global voice biomarker market to grow from $3.35 billion in 2025 to $7.41 billion by 2030, as similar automated diagnostic tools see broader integration into routine medical practice overseas.
Voice Biomarker Detection Frequently Asked Questions
How does the AI distinguish dementia patients from healthy individuals?
The system analyzes both what a patient says and how they say it by combining OpenAI’s Whisper model with the GPT-5.2 large language model. It evaluates speaking speed, word-recall pauses, vocabulary, and sentence structure to achieve an overall classification accuracy of 90.14 percent.
Which organizations are conducting the current clinical trials?
Naver Cloud is conducting clinical trials in partnership with Emocog and Seoul Boramae Hospital to test the effectiveness of detecting cognitive decline through telephone interactions.
What is the projected value of the global voice biomarker market?
The Business Research Company projected the global voice biomarker market to expand from $3.35 billion in 2025 to $7.41 billion by 2030.
Keep reading