Scholarlink

<p> The popularity of biometric identification has grown due to its precision and non-invasive nature. Photoplethysmography (PPG), a technique for monitoring blood volume changes in peripheral tissues, has emerged as a promising modality for identity verification. In this study, we analyze PPG signals collected from 100 individuals under controlled, stable conditions to evaluate the feasibility of using concise, feature-driven machine learning (ML) models for biometric authentication. Our primary objective is to streamline feature extraction for efficient and rapid identity verification, particularly on portable or resource-constrained devices. We applied ML techniques, including support vector machines (SVM), logistic regression, and gradient boosting, supported by hyperparameter tuning using auxiliary AI tools on a GPU. Through targeted feature engineering and data refinement, our model achieved a classification accuracy of 99.64% using only a single cardiac cycle. This result highlights the potential of PPG-based verification in constrained environments. However, the findings should be interpreted within the context of the study’s limitations: the homogeneous sample, the controlled acquisition environment, and the lack of testing under real-world variability. Accordingly, this work is positioned not as a universally robust solution, but as a proof-of-concept for lightweight, context-dependent biometric verification using concise physiological features.</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

Concise PPG features for lightweight identity verification

  • 期刊论文

  • 发表日期：

    2026年2月6日

查看全文
  • Danyal Shahmirzadi,
  • Wen Fong Wang

摘要

The popularity of biometric identification has grown due to its precision and non-invasive nature. Photoplethysmography (PPG), a technique for monitoring blood volume changes in peripheral tissues, has emerged as a promising modality for identity verification. In this study, we analyze PPG signals collected from 100 individuals under controlled, stable conditions to evaluate the feasibility of using concise, feature-driven machine learning (ML) models for biometric authentication. Our primary objective is to streamline feature extraction for efficient and rapid identity verification, particularly on portable or resource-constrained devices. We applied ML techniques, including support vector machines (SVM), logistic regression, and gradient boosting, supported by hyperparameter tuning using auxiliary AI tools on a GPU. Through targeted feature engineering and data refinement, our model achieved a classification accuracy of 99.64% using only a single cardiac cycle. This result highlights the potential of PPG-based verification in constrained environments. However, the findings should be interpreted within the context of the study’s limitations: the homogeneous sample, the controlled acquisition environment, and the lack of testing under real-world variability. Accordingly, this work is positioned not as a universally robust solution, but as a proof-of-concept for lightweight, context-dependent biometric verification using concise physiological features.