Granted Patent B2

US 12,646,625 · App. 17/729,523 · Granted Jun 2, 2026

Inventors: Julian Gerald Dcruz (Kollam, IN); Pai-Chang Yeh (Zhubei, TW)

G16H50/30G06V10/82G06V20/46G06V40/161G16H15/00G16H30/40

Abstract

Claims (38)

1 . A computer-implemented method for non-invasive estimation of a user's pulse rate using only a video camera, the method comprising:

b. capturing, via a monocular RGB video camera operating at ≥30 fps under ambient lighting, a sequence of facial video frames of a subject;

c. detecting candidate skin regions of interest (ROIs) by:

i. segmenting facial landmarks and regions using geometric heuristics or machine learning; and

ii. computing a temporal iPPG signal quality metric to select stable pulsatile ROIs;

d. extracting iPPG signals and temporal features from the retained ROIs, including pulse waveform amplitude, harmonics, and temporal SNR;

e. inferring an OCT-variation feature map for each ROI by executing a trained convolutional neural network on the facial video frames, wherein said CNN has been trained to map RGB-video time series to features that emulate structural or coherence-depth patterns characteristic of OCT imaging, without using interferometric OCT hardware during inference;

f. combining the iPPG features and the OCT-variation feature map to generate a fused feature vector;

g. predicting, via a trained model, a pulse rate value from the fused feature vector; and

h. outputting, via a display or remote system, the pulse rate together with an interpretability cue identifying the ROI used for inference,

Wherein the method is performed by one or more processors executing instructions stored on non-transitory memory, and no physical OCT sensor is used in any step of the method.

2 . The method of claim 1 , wherein the at least one obtained result indicates the pulse rate level of the user within at least one of a healthy range, a caution range, or an abnormal range category.

3 . The method of claim 1 , wherein the at least one obtained result is further analyzed to predict one or more potential health conditions of the user based on the at least one obtained result.

4 . The method of claim 1 , wherein the one or more photoplethysmography imaging (iPPG) variations and the one or more optical coherence tomography (OCT) variations are correlated by the image-based physiological monitoring model to improve accuracy in obtaining the at least one result using the convolutional neural network.

5 . The method of claim 1 , wherein the at least one image- based physiological monitoring model utilizes an artificial intelligence (AI) technique, a deep learning technique, or a trained machine-learning classifier to obtain the at least one result.

6 . The method of claim 1 , wherein the at least one image- based physiological monitoring model comprises a convolutional neural network (CNN) algorithm implemented in software to obtain the at least one result.

7 . The method of claim 1 , wherein processing each frame further comprises applying noise reduction to the frame and performing one or more data augmentations on the noise-reduced frame.

8 . The method of claim 1 , wherein extracting the one or more facial regions and processing the regions of interest further comprises identifying a plurality of correlated sub-regions within each region of interest for analysis.

9 . The method of claim 1 , wherein the one or more cameras include an infrared thermal camera, and the method further comprises determining a body temperature of the user based on infrared imaging data captured by the infrared thermal camera.

10 . The method of claim 1 , wherein the at least one obtained result further comprises an estimation of the blood oxygen saturation (SpO 2 ) level of the user, determined from the real-time video of the user.

11 . The method of claim 1 , wherein the processor is further configured to display the at least one obtained result on a user interface, the user interface presenting the pulse rate level of the user in at least one of a healthy range, a caution range, or an abnormal range, and further displaying an indication of one or more potential diseases predicted based on the at least one obtained result.

12 . A non-contact and non-invasive apparatus for monitoring health characteristics of a user, the apparatus comprising:

a. one or more sensors configured to non-invasively collect data from the user, including at least one camera sensor for capturing real-time video of the user;

b. a processor operatively connected to the one or more sensors, the processor configured to:

c. process the real-time video to extract one or more facial regions in each video frame and identify one or more regions of interest from the facial regions;

d. apply at least one image-based physiological monitoring model to the one or more regions of interest to determine at least one health-related result for the user, wherein the model uses photoplethysmography imaging data and generates optical coherence tomography (OCT) data via a CNN, without requiring any OCT imaging hardware, to compute the at least one health-related result indicative of the user's pulse rate; and

e. optionally compare the captured imaging data or the at least one result with standardized physiological information to identify any anomalies in the user's health characteristics;

f. a communication interface configured to transmit the at least one result to an AI database and to retrieve information from at least one external information source containing standardized physiological data; and

g. an information analysis system in communication with the processor, the information analysis system being configured to utilize data from the at least one external information source for assisting in the determination of the user's health characteristics.

13 . The non-contact and non-invasive apparatus of claim 12 , wherein the one or more sensors comprise one or more of: an infrared thermal detector, a temperature sensor, and a camera sensor.

14 . The non-contact and non-invasive apparatus of claim 12 , wherein:

a. the data collected by the one or more sensors comprises imaging data associated with the user; and

b. the imaging data associated with the user is compared with standardized information from the at least one information source to detect one or more of the user's health characteristics, including pulse rate, respiration rate, blood oxygen saturation (SpO 2 ), blood pressure, or body temperature.

15 . The non-contact and non-invasive apparatus of claim 12 . wherein:

a. the collected imaging data is fed to an image-based neural network to detect a possible respiratory disease in the user; and

b. the neural network comprises a machine-learning prediction model selected from the group consisting of: a convolutional neural network (CNN), a support vector machine (SVM), an artificial neural network (ANN), a neuro-fuzzy classifier (NFC), or a neuro-wavelet technique (NWT).

16 . The non-contact and non-invasive apparatus of claim 12 . wherein the information analysis system includes an artificial intelligence model configured to analyze information from the at least one external information source for one or more gathered physiological parameters of the user.

17 . The non-contact and non-invasive apparatus of claim 12 . wherein the one or more sensors comprise one or more of: a pulse sensor, a respiration sensor, a temperature sensor, a blood pressure sensor, or a blood oxygen saturation (SpO 2 ) sensor.

Continuity (2)

References Cited (8)