Self-Adaptive Matrix Completion for Heart Rate Estimation from Face Videos under Realistic Conditions
read more
Citations
Algorithmic Principles of Remote PPG
Learning Deep Models for Face Anti-Spoofing: Binary or Auxiliary Supervision
Multimodal Spontaneous Emotion Corpus for Human Behavior Analysis
DeepPhys: Video-Based Physiological Measurement Using Convolutional Attention Networks
Unsupervised skin tissue segmentation for remote photoplethysmography
References
Distributed Optimization and Statistical Learning Via the Alternating Direction Method of Multipliers
The use of fast Fourier transform for the estimation of power spectra: A method based on time averaging over short, modified periodograms
A Singular Value Thresholding Algorithm for Matrix Completion
Exact Matrix Completion via Convex Optimization
Guaranteed Minimum-Rank Solutions of Linear Matrix Equations via Nuclear Norm Minimization
Related Papers (5)
Non-contact, automated cardiac pulse measurements using video imaging and blind source separation.
Remote plethysmographic imaging using ambient light
Frequently Asked Questions (10)
Q2. What are the future works mentioned in the paper "Self-adaptive matrix completion for heart rate estimation from face videos under realistic conditions" ?
Future work guidelines include devising novel feature representations, in alternative to chrominance signals, to further improve the robustness to varying illumination conditions as well as exploiting the feasibility of combining the predicted HR measurements with visual features for spontaneous emotion classification.
Q3. What are the future work guidelines for the proposed approach?
Future work guidelines include devising novel feature representations, in alternative to chrominance signals, to further improve the robustness to varying illumination conditions as well as exploiting the feasibility of combining the predicted HR measurements with visual features for spontaneous emotion classification.
Q4. What is the underlying reason of the chrominance features?
minimizing the rank is a NP-hard problem, and traditionally aconvex surrogate of the rank, the nuclear norm, is used [8]:min EνkEk⇤ + kE−Ck 2 F . (1)Another intrinsic property of the chrominance features is that, since the underlying reason of their oscillation is the internal functioning of the heart, the authors should enforce the estimated chrominance features (those of the low-rank estimated matrix) to be within the heart-rate’s frequency range.
Q5. How many subjects participated in the experiment?
It contains 27 subjects (12 males and 15 females) in total, and each subject participated in two experiments: (i) emotion elicitation and (ii) implicit tagging.
Q6. What is the main disadvantage of using a long analysis window?
The main disadvantage of using a long analysis window is the inability to capture interesting short-time phenomena, such as a sudden HR increase/decrease due to specific emotions [22].
Q7. What is the optimal value for M?
The optimal value for M is obtained from the followingoptimisation problem:min MkM ◦ (F−C)k2F − βkMk1 + µkM− fMk2F , (8)which can be rewritten independently for each entry of M:min mrt2{0,1}(frt − ort) 2mrt + µ(mrt − emrt)2 − βmrt.
Q8. Why does the proposed SAMC achieve higher accuracy than the state-of-the-art?
On this difficult dataset, due to its capacity to select the most reliable chrominance features and ignore the noisy ones, the proposed SAMC achieves significantly higher accuracy than the state-of-the-art.
Q9. How is the low-rank estimated matrix grouped?
On the one hand, since matrix completion problems are usually approached by reducing the matrix rank, the low-rank estimated matrix naturally groups the rows by their linear dependency.
Q10. What are the limitations of the method?
In this work, the authors address the aforementioned limitations by proposing a novel method capable of predicting HR with higher accuracy than the state-of-the-art approaches and of robustly operating on short time sequences in order to detect the instantaneous HR.