Scholarlink

<p>The Habitual activity of a person such as walking, and gym practice had been compared with a ground truth model called sleep activity had been done in their dwelling place to measure SpO<sub>2</sub>. The Methodology proposed would extract statistical data like mean (M), standard deviation (SD), and entropy (E) obtained from face contouring along with personal biological data say heart rate (HR), blood pressure (BP), age, weight, Sex, SpO<sub>2,</sub> etc., Preamble required in this proposed method consists of an experimental setup such as smartphone cameras, pulse oximeter, an event representing person habitual activity. Conventional approaches examine EEG signals, ECG signals, and heart rate Variability from sleep involving expensive sleep labs, host presence in hospitals as well as a survey-based approach to detect sleep apnea. In the proposed system, the <i>first approach</i> utilizes supervised ML models by considering face features and biological data with RMSE as validation metrics giving better results in walking as habitual activity across compared 10 machine learning models. The s<i>econd approach</i> compares the face image at rest and the face image after each habitual activity among five performance metrics scores such as Jaccard’s Index (JI), Mathew Correlation Coefficient (MCC), Accuracy (A), Precision (P), Recall (R) to monitor heart wellbeing by exhibiting the excellent result with walk as habitual activity and SpO<sub>2</sub> as prediction parameter against face features. The resultant of the two approaches ensures superior performance for walking as a habitual activity with JI = 95.92, MCC = 93.64%, A = 97.19%, P = 98.40%, and R = 97.43% respectively.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

SpO2 Prediction Using Machine Learning Approach by Comparing Habitual Activity and Sleep Stages with Improved Accuracy for Walk Inside Dwelling Environment

  • 期刊论文

  • 发表日期：

    2025年6月25日

查看全文
  • K. Krishnakumar,
  • S Latha

摘要

The Habitual activity of a person such as walking, and gym practice had been compared with a ground truth model called sleep activity had been done in their dwelling place to measure SpO2. The Methodology proposed would extract statistical data like mean (M), standard deviation (SD), and entropy (E) obtained from face contouring along with personal biological data say heart rate (HR), blood pressure (BP), age, weight, Sex, SpO2, etc., Preamble required in this proposed method consists of an experimental setup such as smartphone cameras, pulse oximeter, an event representing person habitual activity. Conventional approaches examine EEG signals, ECG signals, and heart rate Variability from sleep involving expensive sleep labs, host presence in hospitals as well as a survey-based approach to detect sleep apnea. In the proposed system, the first approach utilizes supervised ML models by considering face features and biological data with RMSE as validation metrics giving better results in walking as habitual activity across compared 10 machine learning models. The second approach compares the face image at rest and the face image after each habitual activity among five performance metrics scores such as Jaccard’s Index (JI), Mathew Correlation Coefficient (MCC), Accuracy (A), Precision (P), Recall (R) to monitor heart wellbeing by exhibiting the excellent result with walk as habitual activity and SpO2 as prediction parameter against face features. The resultant of the two approaches ensures superior performance for walking as a habitual activity with JI = 95.92, MCC = 93.64%, A = 97.19%, P = 98.40%, and R = 97.43% respectively.