Estimating and Analyzing Daily Sleep Quality Using Deep Learning with Multiple Factors

2023 Fiscal Year Final Research Report

• Project/Area Number: 22K19832
• Research Category: Grant-in-Aid for Challenging Research (Exploratory)
• Allocation Type: Multi-year Fund
• Review Section: Medium-sized Section 62:Applied informatics and related fields
• Research Institution: Osaka University
• Principal Investigator: Fukui Ken-ichi 大阪大学, 産業科学研究所, 准教授 (80418772)
• Co-Investigator(Kenkyū-buntansha): 加藤 隆史 大阪大学, 大学院歯学研究科, 教授 (50367520)
• Project Period (FY): 2022-06-30 – 2024-03-31
• Keywords: 睡眠 / 音響 / 深層学習 / ドメイン適応 / マルチモーダル

Outline of Final Research Achievements

In this study, we developed a deep learning model to estimate the quality of daily sleep based on sounds, enabling non-contact and easy measurement. Over a continuous four-week period, we collected data on sleep sounds, heart rate, activity levels during the day, bedroom temperature, humidity, and light levels, and administered surveys about sleep quality, health, and the environment before bed and after waking. 1. 56 additional subjects in their 40s to 60s were recruited, bringing the total to 385 participants. 2. To mitigate individual and environmental differences in sound features, we developed a sleep evaluation method based on domain adaptation and verified its effectiveness using our home database. 3. We proposed two types of deep learning models capable of sleep evaluation, incorporating physical and environmental factors in addition to sleep sounds, revealing the main factors affecting sleep for different age groups.

Free Research Field

知能情報，機械学習，応用情報

Academic Significance and Societal Importance of the Research Achievements

本研究は非接触かつ簡便に計測可能な音響に基づく新しい睡眠評価法を提案し，日常の睡眠評価の高精度化に貢献した．本研究により開発した深層学習に基づく睡眠評価モデルは，睡眠音特徴の個人差や環境差に対応し広範囲に利用可能であり，また個人毎に身体・環境要因などと統合して睡眠影響因子を特定が可能である．これにより，将来的に個別化された具体的な睡眠改善策の提供が可能となり，日常の睡眠の質改善により公衆衛生の向上，メンタルヘルス問題などの予防に貢献するものと考える．