نوع مقاله : Original Article(s)
نویسندگان
1 دکتری، گروه پردازش علائم زیستی، پژوهشکده پردازش سیگنال، پژوهشگاه توسعه فناوریهای پیشرفته، تهران، ایران
2 کارشناسی ارشد، گروه پردازش علائم زیستی، پژوهشکده پردازش سیگنال، پژوهشگاه توسعه فناوریهای پیشرفته، تهران، ایران
چکیده
تازه های تحقیق
سارا پورمحمدی: PubMed, Google Scholar
کیان شاهی: Google Scholar
کلیدواژهها
موضوعات
عنوان مقاله [English]
نویسندگان [English]
Background: Using remote photoplethysmography (rPPG) technology, monitoring heart rate (HR) has become possible without physical contact. Over the past decades, methods have been developed to estimate rPPG signals and heart rates using video frames. Recently, deep learning techniques have also been applied in this field, showing promising performance.
Methods: In this study, a deep neural network (MTTS-CAN) along with a face detection algorithm (MediaPipe) was used to estimate heart rate remotely from videos of individuals in a public dataset (PURE) and a local dataset (Stroop).
Findings: The implementation results on the PURE dataset are comparable to those published in articles (MAE: 7.72 bpm). Moreover, the results on the local dataset are also acceptable (MAE: 5.53 bpm).
Conclusion: This paper presents an acceptable accuracy for non-contact heart rate estimation. Additionally, the results indicate that the proposed method is not dependent on a specific dataset and has produced satisfactory results in a local dataset compared to the benchmark dataset.
کلیدواژهها [English]
Deniz Cyli N. rPPG based heart rate estimation using deep learning [Thesis]. Istanbul: Marmara University; 2021.