Physiological parameter data calculation method based on facial image

Abstract

The invention relates to the technical field of physiological parameter data calculation, and discloses a physiological parameter data calculation method based on a facial image, which mainly comprises the following steps: S1, face detection and positioning: firstly, the position of a face needs to be detected in an image or a video, and the step is the basis of subsequent processing; s2, face image preprocessing: rotating, zooming and the like are performed on the detected face, so that the face is aligned with the standard template, and standardization of the face image is ensured; s3, skin color change capture: the color change of the facial skin caused by heartbeat is captured by using a camera, and the change is mainly caused by periodic change of blood flow in capillaries due to heartbeat, so that the light absorption amount of the skin is influenced; the method has the advantages that the error recognition rate is reduced, the problems existing in physiological parameter data calculation based on the face image can be effectively solved, and the calculation accuracy and stability are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of physiological parameter data calculation, and in particular to a physiological parameter data calculation method based on facial images. Background Art

[0002] By capturing the changes in skin color on the surface of the face caused by the heartbeat, it is possible to measure physiological indicators such as heart rate and respiratory rate. The calculation of physiological parameter data based on facial images has wide applications and important significance in many fields such as health monitoring, human-computer interaction, security, sports fitness and emotion recognition.

[0003] Chinese patent CN118892316A proposes a physiological parameter detection method, a vehicle, a storage medium and a computer program product. The physiological parameter detection method is applied to a vehicle, and the vehicle includes an image acquisition device and a heat acquisition device. The physiological parameter detection method of the present application specifically includes: capturing multiple facial image data containing a driver through the image acquisition device, and capturing multiple thermal image data containing the driver through the thermal acquisition device; processing the multiple facial image data to obtain facial pixel parameter change values ​​corresponding to the driver, and determining the first physiological parameter corresponding to the driver according to the facial pixel parameter change values; processing the multiple thermal image data to obtain the second physiological parameter corresponding to the driver; and obtaining the target physiological parameter corresponding to the driver based on the first physiological parameter and the second physiological parameter, thereby achieving the technical effect of enabling the vehicle to accurately obtain the physiological parameters of the driver.

[0004] However, the above scheme is prone to problems such as insufficient lighting and uneven lighting, which will lead to a decrease in the quality of facial images and affect the accuracy of physiological parameter calculation. The face is blocked (such as masks, hats, etc.) and some facial features cannot be recognized, affecting the accurate calculation of physiological parameters. In complex scenarios, such as large-scale gatherings of people, changes in light, etc., the facial recognition system is easily disturbed, resulting in missed recognition or incorrect recognition. Summary of the invention

[0005] The purpose of the present invention is to provide a method for calculating physiological parameter data based on facial images in order to solve the above problems, thereby solving the problems mentioned in the background technology.

[0006] In order to solve the above problems, the present invention provides a technical solution:

[0007] A method for calculating physiological parameter data based on facial images, comprising the following steps:

[0008] S1: Face detection and positioning: First, it is necessary to detect the face and position in the image or video. This step is the basis for subsequent processing;

[0009] S2: Facial image preprocessing: rotating and scaling the detected faces to align them with the standard template to ensure the standardization of the facial images;

[0010] S3: Capture of skin color changes: Use a camera to capture the color changes of facial skin caused by heartbeats. This change is mainly due to the periodic changes in blood flow in capillaries caused by heartbeats, which affects the amount of light absorbed by the skin.

[0011] S4: Use high-quality camera equipment: Choose high-quality camera equipment to ensure that the facial images collected are clear and evenly illuminated, and reduce errors caused by equipment differences;

[0012] S5: Multimodal fusion: Fusion of facial images with other biometric features (such as fingerprints, irises, etc.);

[0013] S6: Multi-angle acquisition: Use multiple cameras to capture facial images from different angles;

[0014] S7: Data encryption and privacy protection: Use encryption algorithms to encrypt facial images, and establish a sound data security management system and privacy policy.

[0015] As a preferred embodiment of the present invention, the method further comprises the following steps:

[0016] S8: Restore BVP signal: By analyzing the facial color change signal, restore the brain blood volume pulse (BVP) signal, which contains information such as heart rate and respiratory rate;

[0017] S9: Physiological parameter calculation: By measuring basic physiological parameters such as blood pressure, heart rate, body temperature, blood oxygen saturation, etc., a preliminary assessment of the individual's health status can be made.

[0018] As a preferred solution of the present invention, the facial image preprocessing in S2 mainly includes the following steps:

[0019] S21: Grayscale transformation: Digital images exist in the form of bitmaps on computers, and the brightness of each pixel is called grayscale;

[0020] S21: Threshold segmentation: By comparing the grayscale difference between the target area and the background area, a threshold is determined for segmentation;

[0021] S21: Alignment and cropping: Align and crop the image to ensure that the face is in the same position in the image and remove unnecessary background information;

[0022] S21: Geometric correction: Correct image deformation caused by changes in shooting angle or posture;

[0023] S21: Filtering and sharpening: Enhance the edge information and details of the image through filtering and sharpening.

[0024] As a preferred solution of the present invention, the capturing according to the skin color change in S3 mainly comprises the following steps:

[0025] S31: Image processing: The collected photos will go through a series of image processing steps, including denoising, contrast enhancement, color correction, etc.

[0026] S31: Color correction technology: Through color correction technology, the influence of changes in lighting conditions on skin color detection can be eliminated;

[0027] S31: Multi-color space selection: Select the appropriate color space for skin detection;

[0028] S31: Utilization of texture information: Texture information is introduced into skin detection, and the texture information of the image is extracted through the gray-level co-occurrence matrix. The texture information is used to further filter the pixel areas whose color information is skin and whose texture information is non-skin.

[0029] As a preferred solution of the present invention, the calculation according to the physiological parameters in S9 mainly includes the following steps:

[0030] S91: Heart rate calculation: The heart rate can be obtained by analyzing the frequency domain characteristics of the BVP signal;

[0031] S92: Respiration rate calculation: The respiratory rate can be calculated by analyzing the frequency spectrum characteristics of the BVP signal;

[0032] S93: Blood oxygen content (SpO2) estimation: The blood oxygen content is estimated by using the ratio of the DC and AC quantities of signals from different color channels;

[0033] S94: Heartbeat variability analysis: Physiological quantities such as heartbeat variability can be obtained by detecting the timing peak value of the BVP signal.

[0034] As a preferred solution of the present invention, the recovery of the BVP signal in S8 and the detection of the psychological reaction of the human body when facing different psychological pressures through data analysis of the BVP signal can reflect the stability of the human autonomic nervous system and the regulatory ability of the heart.

[0035] As a preferred solution of the present invention, the multimodal fusion in S5 mainly includes the following steps:

[0036] S51: Multimodal data acquisition: In addition to facial images, other physiological parameter data such as heart rate, respiratory rate, etc. need to be acquired;

[0037] S52: Feature extraction: Use deep learning models to extract features from facial images. The model extracts facial key points (such as the position and shape of eyes, nose, mouth, etc.), as well as facial expressions and movement features;

[0038] S53: Multimodal data fusion: Fusion of extracted facial image features with other physiological parameter data;

[0039] S54: Parameter calculation: Utilize the fused multimodal data to calculate physiological parameters through machine learning models.

[0040] The beneficial effects of the present invention are: by setting up and using high-quality camera equipment, multimodal fusion and multi-angle acquisition, adopting more advanced image processing algorithms and deep learning technology, the extraction and matching accuracy of facial features can be improved, and the misrecognition rate can be reduced. It can effectively solve the problems existing in the calculation of physiological parameter data based on facial images and improve the accuracy and stability of the calculation. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] For ease of explanation, the present invention is described in detail with reference to the following specific implementations and the accompanying drawings.

[0042] Figure 1 It is a flow chart of physiological parameter data calculation of the present invention. DETAILED DESCRIPTION

[0043] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0044] Embodiment 1:

[0045] See also Figure 1 The present invention provides a technical solution: a method for calculating physiological parameter data based on facial images, which mainly includes the following steps:

[0046] S1: Face detection and positioning: First, the face position needs to be detected in the image or video. This step is the basis for subsequent processing;

[0047] S2: Facial image preprocessing: rotating and scaling the detected faces to align them with the standard template to ensure the standardization of the facial images;

[0048] S3: Capture of skin color changes: Use a camera to capture the color changes of facial skin caused by heartbeats. This change is mainly due to the periodic changes in blood flow in capillaries caused by heartbeats, which affects the amount of light absorbed by the skin.

[0049] S4: Use high-quality camera equipment: Choose high-quality camera equipment to ensure that the facial images collected are clear and evenly illuminated, and reduce errors caused by equipment differences;

[0050] S5: Multimodal fusion: Fusion of facial images with other biometric features (such as fingerprints, irises, etc.) to improve recognition accuracy and robustness;

[0051] S6: Multi-angle acquisition: Use multiple cameras to collect facial images from different angles to increase the success rate of recognition;

[0052] S7: Data encryption and privacy protection: Use encryption algorithms to encrypt facial images, and establish a sound data security management system and privacy policy.

[0053] It can effectively solve the problems existing in the calculation of physiological parameter data based on facial images and improve the accuracy and stability of the calculation.

[0054] Embodiment 2:

[0055] The following steps are also included:

[0056] S8: Restore BVP signal: By analyzing the facial color change signal, restore the brain blood volume pulse (BVP) signal, which contains information such as heart rate and respiratory rate;

[0057] S9: Physiological parameter calculation: By measuring basic physiological parameters such as blood pressure, heart rate, body temperature, blood oxygen saturation, etc., a preliminary assessment of the individual's health status can be made.

[0058] Together, they constitute the process of calculating physiological parameter data based on facial images, which can effectively extract physiological parameters such as heart rate, respiratory rate, and blood oxygen content from facial images.

[0059] Embodiment 3:

[0060] According to S2, facial image preprocessing mainly includes the following steps:

[0061] S21: Grayscale transformation: Digital images exist in the form of bitmaps on computers. The brightness of each pixel is called grayscale. Grayscale transformation is to transform the grayscale value of each pixel on the image according to a certain function. Commonly used methods include linear transformation, piecewise linear transformation and nonlinear grayscale transformation. These methods can increase the dynamic range of the image, expand the contrast, make the image clearer, and make the features more obvious.

[0062] S22: Threshold segmentation: By comparing the grayscale difference between the target area and the background area, a threshold is determined for segmentation; if the target area is large enough and has a significant grayscale difference with the background, the threshold can be determined by histogram equalization. This method can effectively divide the image into foreground and background.

[0063] S23: Alignment and cropping: Align and crop the image to ensure that the face is positioned consistently in the image and remove unnecessary background information; this helps improve the accuracy and efficiency of feature extraction.

[0064] S24: Geometric correction: Correct image deformation caused by changes in shooting angle or posture; ensure that the geometric shape of the image is correct.

[0065] S25: Filtering and sharpening: Through filtering and sharpening processing, the edge information and details of the image are enhanced to improve the image quality.

[0066] In order to reduce noise, enhance contrast, correct light and geometric deformation, and extract high-quality facial image features, it provides a reliable data basis for subsequent physiological parameter calculations.

[0067] Embodiment 4:

[0068] Capturing skin color changes in S3 mainly includes the following steps:

[0069] S31: Image processing: The collected photos will go through a series of image processing steps, including denoising, contrast enhancement, color correction, etc., to enhance the photo quality and provide more accurate input for subsequent analysis.

[0070] S31: Color correction technology: Color correction technology can eliminate the impact of changes in lighting conditions on skin color detection; use the "gray world" algorithm to correct color changes caused by non-standard lighting, effectively improving the positive detection rate.

[0071] S31: Multi-color space selection: Select a suitable color space for skin detection; use CIE Lab color space for histogram analysis to determine whether there is color cast in the image to be detected, and perform corresponding color correction.

[0072] S31: Utilization of texture information: Texture information is introduced into skin detection. The texture information of the image is extracted through the gray-level co-occurrence matrix. The texture information is used to further filter the pixel areas where the color information is skin but the texture information is non-skin, so as to improve the detection performance.

[0073] It can effectively improve the accuracy and reliability of capturing skin color changes, thereby better serving skin detection and diagnosis.

[0074] Embodiment 5:

[0075] The calculation of physiological parameters in S9 mainly includes the following steps:

[0076] S91: Heart rate calculation: The heart rate can be obtained by analyzing the frequency domain characteristics of the BVP signal;

[0077] S92: Respiration rate calculation: The respiratory rate can be calculated by analyzing the frequency spectrum characteristics of the BVP signal;

[0078] S93: Blood oxygen content (SpO2) estimation: The blood oxygen content is estimated by using the ratio of the DC and AC quantities of signals from different color channels;

[0079] S94: Heartbeat variability analysis: Physiological quantities such as heartbeat variability can be obtained by detecting the timing peak value of the BVP signal.

[0080] These steps together constitute the process of calculating physiological parameter data based on facial images, which can effectively extract physiological parameters such as heart rate, respiratory rate, blood oxygen content, etc. from facial images.

[0081] Embodiment 6:

[0082] According to the restored BVP signal in S8, through the data analysis of the BVP signal, the psychological reaction of the human body when facing different psychological pressures is detected, which can reflect the stability of the human autonomic nervous system and the regulatory ability of the heart.

[0083] Embodiment 7:

[0084] According to S5, multimodal fusion mainly includes the following steps:

[0085] S51: Multimodal data acquisition: In addition to facial images, other physiological parameter data such as heart rate, respiratory rate, etc. need to be acquired; ensure that the image clarity and resolution are high enough for subsequent processing.

[0086] S52: Feature extraction: Use deep learning models to extract features from facial images. The model extracts facial key points (such as the position and shape of eyes, nose, mouth, etc.), as well as facial expressions and movement features;

[0087] S53: Multimodal data fusion: Fuse the extracted facial image features with other physiological parameter data; this can be achieved through methods such as feature vector concatenation and weighted averaging. The fused data will contain richer information and help to estimate physiological parameters more accurately.

[0088] S54: Parameter calculation: Using the fused multimodal data, physiological parameters are calculated through machine learning models. Emotional states can be estimated by analyzing changes in facial expressions, and stress levels can be assessed by combining changes in heart rate.

[0089] It can realize accurate calculation of physiological parameter data based on facial images and be applied in various practical scenarios to improve system reliability and user experience.

[0090] Embodiment 8:

[0091] The technology measures heart rate by capturing the skin color changes on the face surface due to the heartbeat. When the heart beats, the blood flow in the capillaries will change periodically, resulting in periodic changes in skin color. By capturing this color change with a camera, the brain blood volume pulse (BVP) signal can be restored to calculate the heart rate. In the process of feature extraction and analysis, the deep learning algorithm learns the relationship between facial features and physiological parameters by training a large amount of facial data, thereby predicting physiological parameters. It has important application value in the field of health monitoring and can monitor physiological parameters such as heart rate and respiratory rate in real time. Through the analysis of physiological quantities such as heart rate variability, it can be used for sleep quality monitoring and stress detection.

[0092] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for calculating physiological parameter data based on facial images, characterized in that: The main steps include: S1: Face detection and positioning: First, the position of the face needs to be measured in the image or video. This step is the basis for subsequent processing; S2: Facial image preprocessing: rotating and scaling the detected faces to align them with the standard template to ensure the standardization of the facial images; S3: Capture of skin color changes: Use a camera to capture the color changes of facial skin caused by heartbeats. This change is mainly due to the periodic changes in blood flow in capillaries caused by heartbeats, which affects the amount of light absorbed by the skin. S4: Use high-quality camera equipment: Choose high-quality camera equipment to ensure that the facial images collected are clear and evenly illuminated, and reduce errors caused by equipment differences; S5: Multimodal fusion: Fusion of facial images with other biometric features (such as fingerprints, irises, etc.); S6: Multi-angle acquisition: Use multiple cameras to capture facial images from different angles; S7: Data encryption and privacy protection: Use encryption algorithms to encrypt facial images, and establish a sound data security management system and privacy policy.

2. The method for calculating physiological parameter data based on facial images according to claim 1, characterized in that: The method further comprises the following steps: S8: Restore BVP signal: By analyzing the facial color change signal, restore the brain blood volume pulse (BVP) signal, which contains information such as heart rate and respiratory rate; S9: Physiological parameter calculation: By measuring basic physiological parameters such as blood pressure, heart rate, body temperature, blood oxygen saturation, etc., a preliminary assessment of the individual's health status can be made.

3. The method for calculating physiological parameter data based on facial images according to claim 1, characterized in that: The facial image preprocessing according to S2 mainly includes the following steps: S21: Grayscale transformation: Digital images exist in the form of bitmaps on computers, and the brightness of each pixel is called grayscale; S22: Threshold segmentation: By comparing the grayscale difference between the target area and the background area, a threshold is determined for segmentation; S23: Alignment and cropping: Align and crop the image to ensure that the face is in the same position in the image and remove unnecessary background information; S24: Geometric correction: Correcting image deformation caused by changes in shooting angle or posture; S25: Filtering and sharpening: Enhance the edge information and details of the image through filtering and sharpening.

4. The method for calculating physiological parameter data based on facial images according to claim 1, characterized in that: The capturing of skin color changes in S3 mainly includes the following steps: S31: Image processing: The collected photos will go through a series of image processing steps, including denoising, contrast enhancement, color correction, etc. S32: Color correction technology: Through color correction technology, the influence of changes in lighting conditions on skin color detection can be eliminated; S33: Multi-color space selection: Select the appropriate color space for skin detection; S34: Utilization of texture information: Texture information is introduced into skin detection, and the texture information of the image is extracted through the gray-level co-occurrence matrix. The texture information is used to further filter the pixel areas whose color information is skin and whose texture information is non-skin.

5. The method for calculating physiological parameter data based on facial images according to claim 2, characterized in that: The calculation according to the physiological parameters in S9 mainly includes the following steps: S91: Heart rate calculation: The heart rate can be obtained by analyzing the frequency domain characteristics of the BVP signal; S92: Respiration rate calculation: The respiratory rate can be calculated by analyzing the frequency spectrum characteristics of the BVP signal; S93: Blood oxygen content (SpO2) estimation: The blood oxygen content is estimated by using the ratio of the DC and AC quantities of signals from different color channels; S94: Heartbeat variability analysis: Physiological quantities such as heartbeat variability can be obtained by detecting the timing peak value of the BVP signal.

6. The method for calculating physiological parameter data based on facial images according to claim 2, characterized in that: The recovery of the BVP signal in S8 and the detection of the psychological reaction of the human body when facing different psychological pressures through data analysis of the BVP signal can reflect the stability of the human autonomic nervous system and the regulatory ability of the heart.

7. The method for calculating physiological parameter data based on facial images according to claim 1, characterized in that: The multimodal fusion according to S5 mainly includes the following steps: S51: Multimodal data acquisition: In addition to facial images, other physiological parameter data such as heart rate, respiratory rate, etc. need to be acquired; S52: Feature extraction: Use deep learning models to extract features from facial images. The model extracts facial key points (such as the position and shape of eyes, nose, mouth, etc.), as well as facial expressions and movement features; S53: Multimodal data fusion: Fusion of extracted facial image features with other physiological parameter data; S54: Parameter calculation: Utilize the fused multimodal data to calculate physiological parameters through machine learning models.

Images