Asleep leads the innovation in sleep research

Asleep is leading the innovation in sleep by utilizing AI technology to enable everyone to understand and improve their own sleep.
Sleep tracking: Why breathing sounds?
Data incoming...
The activation levels of the autonomic nervous system and the motor nervous system vary according to the sleep stage. Analyzing breathing sounds allows us to identify the activation levels of each nervous system, thereby enabling us to deduce the sleep stage inversely.

The state of the autonomic nervous system can be understood through the rhythm, cycle, and pattern of breathing. Breathing also provides crucial clues when assessing the state of the motor nervous system.

The tone and depth of breathing allow us to track the state of the motor nervous system. Furthermore, by understanding how breathing sounds and patterns change, it is possible to detect conditions like sleep apnea or hypoventilation during sleep.
Sleep sound and sleep stage ↗️
Autonomic nervous system
Motor nervous system
Detect disease
Polysomnography(PSG)
Heartbeat Tracking(Smart Watch, Ring)
Breathing sound

(Asleep AI)
Rhythm and Cycle
Tone and depth
Sound and pattern
Distinguish the breathing like AI
Compare the apnea and normal breathing detected by Asleep AI.
More breathing samples↗️
01 Normal Breathing
Normal breathing sound without any event.
02 Apnea
No breath sound can be heard.
03 Hypopnea
The breathing sounds get weaker and the pattern gets unstable.
These breathing sounds are used solely for the purpose of understanding, and the actual AI does not listen to any sounds at all.
Asleep AI Collects Data Safely ↓️

Breathing, the most accurate tracker

In accuracy benchmark studies including wearable and nearable sleep trackers from big tech companies, Asleep AI showed the highest accuracy as of April 2024.
Sleep tracker comparison study (SCI) ↗
6,090
Hospital PSG Study Count
The AI was trained with polysomnography (PSG) data that had been cross-read by sleep technicians and sleep specialists. This amounts to approximately 25 years of learning data (based on one night per case, five days a week).
Completed accuracy validation for multi-ethnic groups
Completed accuracy validation across different BMIs and genders
400
Home Environment PSG Study Count
The AI also learned from polysomnography data conducted in personalized home environments, not just controlled hospital settings. This ensures accurate sleep trackings even in unpredictable home environments.
Completed learning of noises occurring in home environments
2,201,145
Service Sleep Data Study Count
Through the B2C service SleepRoutine, it has successfully acquired sleep tracking data in various environments and has gone through a fine-tuning process for actual API service.
Completed tests with low-quality microphones
Completed tests with built-in microphones across all smartphone models
2m Completed accuracy tests within a 2-meter radius

Transforming sleep science through global collaboration

“수면 의학과 AI의 결합은 수면 진단을 발전시키고, 나아가 수면의 질을 개선할 수 있는 잠재력이 충분합니다”
Clete A. Kushida
Ph.D, Medical director of the Stanford Sleep Disorders Clinic
“가정에서 편안하게 잠을 자는 동안 수면 단계를 비접촉방식으로 측정할 수 있는 기술은 그 자체로 의의가 큽니다. 향후가 기대됩니다”
김정훈
분당서울대 이비인후과 교수
“We think technology can help make it easier to do important things during the day and night so they can be their Best Slept Self”
John Lopos
CEO, National Sleep Foundation
“수면 관련 질환으로 이어질 수 있는 환자들을 조기에 진단하고, 적극적으로 치료를 받는 데 도움을 줄 수 있을 것으로 기대하고 있습니다”
유인영
분당서울대 이비인후과 교수
“불면증 환자에게 디지털로 교육하면 행동이 바뀌고, 이후에 증상이 악화되지 않고 유지가 되는 것을 목격했습니다. 시장이 충분히 성장하지는 않았지만, 디지털 치료기기는 충분히 기대감을 가질 만 합니다”
신재용 교수
연세대학교 의과대학
"The combination of sleep medicine and AI has the potential to advance sleep diagnostics and, furthermore, improve the quality of sleep."
Clete A. Kushida
Ph.D, Medical director of the Stanford Sleep Disorders Clinic
"The technology that allows for the non-contact tracking of sleep stages while comfortably sleeping at home is significant in itself. The future looks promising."
김정훈
분당서울대 이비인후과 교수
“We think technology can help make it easier to do important things during the day and night so they can be their Best Slept Self”
John Lopos
CEO, National Sleep Foundation
"We are hopeful that it will help in the early diagnosis of patients who could develop sleep-related disorders and encourage them to seek active treatment."
유인영
분당서울대 이비인후과 교수
"We are hopeful that it will help in the early diagnosis of patients who could develop sleep-related disorders and encourage them to seek active treatment."
신재용 교수
연세대학교 의과대학

Asleep AI collects data safely

Asleep AI is an AI that analyzes images. It uses breathing sounds, but it doesn't actually listen to the sounds. It employs Mel-spectrogram technology, which converts the voice into images that retain only the information necessary for sleep analysis. This process is done entirely on the user's device (locally) before the images are transmitted, making it impossible to access the breathing sounds or any sounds.
Safely Managed Asleep Data ↗
Never listens to sounds
Uses an image AI that interprets sleep stages through Mel-spectrograms.
Cannot be traced-back
Images are implemented in a flat format differently from voice, making it impossible to convert images back.
Automatic filtering
All sounds other than breathing are classified as noise and are not mapped onto the image.

Only 5% of information remains for analysis

A single image converted from a breathing sound
Before being transmitted to the server, the conversion into an image takes place within the user's mobile phone.

After the image processing, all data including the breathing sound is immediately erased, and only a single image is transmitted to the server for AI analysis.
7,680,000 bits
576,000 bits
95% of the information from breathing sounds is discarded
Only 5% of the actual breathing sound's information is necessary for sleep analysis. This 5% of information enables the most accurate non-contact sleep analysis. The information needed for sleep analysis boils down to just six categories:
Rhythm
Cycle
Tone
Depth
Volume
Pattern
(Comparison of file sizes before and after a 30-second analysis, WAV to Mel conversion)
Asleep AI leads innovation
True innovation goes beyond simply developing new technologies; it brings about changes in people's lifestyles.

Innovative ideas sometimes completely break the existing mold, changing many different fields in the process.

Each field must be persuaded in its own way, and Asleep continually attracts attention in the sleep science community by constantly dissecting and presenting the topic of accurate sleep tracking through breathing sounds.
Achievements
74
Patent Applications in Leading Technology Nations such as Korea, the USA, and Japan
The principles, operation methods, and various applications of sleep tracking technology are protected by patents.
Real-time breathing sound sleep analysis is possible only with Asleep AI
Various IoT and sleep tracking AI integration technologies are legally protected
Legal protection for home environment PSG and environment creation devices and programs
Legal protection for sleep analysis and result-displaying UI/UX
22
Record Number of Papers Presented at Sleep Conferences (Single Company)
Various research results for validating the technology and accuracy of Asleep have been presented at multiple sleep conferences, including the World Sleep Congress.
KFDA
Class II Medical Device Approval by the Food and Drug Safety Ministry
A medical auxiliary device equipped with Asleep AI technology has been certified as the first home-use sleep apnea diagnostic auxiliary device in the country, with Class II medical device approval. The apnea diagnosis accuracy of this device surpasses the performance of diagnostic auxiliary devices certified by the FDA
Publications
These are the papers researched and published by Asleep. Read about the sleep tracking AI technology that benefits the world and the various sleep-related research results achieved using AI.
We’re open to collaborating
Asleep's breathing sound-based sleep tracking technology enables sleep-related experiments and validations in various environments that could not be conducted due to accuracy and discomfort issues.

If you need to discuss the possibilities for various intervention studies and validations, please contact Asleep. We will send you cases used for intervention and efficacy verification.