System and method for determining respiration rate from a video
What is disclosed is a system and method for determining respiration rate from a video of a subject. In one embodiment, a video is received comprising plurality of time-sequential image frames of a region of a subject's body. Features of pixels are extracted from that region from each image frame and vectors formed from these features. Each image frame has an associated feature vector. A N×M video matrix of the vectors of length N is constructed such that a total number of columns M in the video matrix correspond to a time duration over which the subject's respiration rate is to be determined. The video matrix is processed to obtain a matrix of eigenvectors where principal axes of variations due to motion associated with respiration are contained in a first few eigenvectors. One eigenvector is selected from the first few eigenvectors. A respiration rate is obtained from the selected eigenvector.
1. A method for determining respiration rate from a video of a subject being monitored for respiratory function, the method comprising:
receiving a video of a subject being monitored for respiratory function, said video comprising a plurality of time-sequential image frames of at least one region of interest of said subject's body where a signal corresponding to respiration can be registered by at least one imaging channel of a video imaging device;
extracting, from said region of interest from said plurality of image frames, features of pixels and forming vectors of length N from said extracted features;
constructing a N×M video matrix of temporally sequential feature vectors, a total number of columns M in said video matrix corresponding to a time duration over which a respiration rate for said subject is to be determined;
processing said video matrix to obtain a matrix of eigenvectors, said eigenvector matrix being such that principal axes of variations due to motion associated with said subject's respiration are contained in a first few eigenvectors;
selecting one eigenvector from said first few eigenvectors; and
extracting a respiration rate for said subject from said selected eigenvector.
2. The method of claim 1 , wherein said feature vector comprises any of: an intensity component, a location component, and a motion component.
3. The method of claim 2 , wherein said video imaging device is a color video camera comprising red, green and blue channels and said intensity component is obtained from any of said channels.
4. The method of claim 2 , wherein said video imaging device is a color video camera comprising red, green and blue channels and said intensity component is obtained from computing an algebraic sum of pixel values from said red, green and blue channels.
5. The method of claim 1 , wherein, in advance of forming said feature vectors, grouping pixels associated with said features and computing their mean values.
6. The method of claim 5 , further comprising spatial filtering said grouped pixels to reduce noise.
7. The method of claim 5 , further comprising amplitude filtering by taking an L2 norm of each row vector and retaining all feature points with an L2 norm that is below a pre-defined maximum value.
8. The method of claim 1 , wherein said video matrix comprises orthonormal eigenvectors obtained by computing mean and covariance matrices using singular value decomposition.
9. The method of claim 8 , wherein, in advance of computing said mean and covariance matrices, computing a derivative of said video matrix.
10. The method of claim 1 , wherein eigenvectors are computed by performing singular value decomposition on said video matrix.
11. The method of claim 10 , wherein, in advance of performing singular value decomposition, computing a derivative of said video matrix.
12. The method of claim 1 , wherein extracting said respiration rate from said selected eigenvector comprises:
generating a power spectral density curve for said selected eigenvector; and
identifying a frequency within said power spectral density curve with a highest signal strength, said identified frequency being said subject's respiration rate.
13. The method of claim 1 , wherein extracting said respiration rate from said selected eigenvector comprises:
computing peak to peak interval between adjacent peaks; and
computing frequency from select group of peak to peak intervals.
14. The method of claim 1 , wherein, in response to said subject's respiration rate not being within a normal range, performing any of: initiating an alert, and signaling a medical professional.
15. The method of claim 1 , further comprising using said respiratory rate to facilitate a determination of an occurrence of any of: Sudden Infant Death Syndrome, Respiratory Distress, Respiratory Failure, and Pulmonary Disease.
16. The method of claim 1 , wherein said video is a streaming video and said respiration rate is determined for said subject in real-time.
17. A system for determining respiration rate from a video of a subject being monitored for respiratory function, the system comprising:
a video imaging device for capturing video comprising a plurality of time-sequential image frames of at least one region of interest of a subject's body where a signal corresponding to respiration can be registered by at least one imaging channel of said video imaging device;
a processor in communication with a memory and executing machine readable instructions for performing:
receiving said video of a subject;
extracting, from said region of interest from said plurality of image frames, features of pixels and forming vectors of length N from said extracted features;
constructing a N×M video matrix of temporally sequential feature vectors, a total number of columns M in said video matrix corresponding to a time duration over which a respiration rate for said subject is to be determined;
processing said video matrix to obtain a matrix of eigenvectors, said eigenvector matrix being such that principal axes of variations due to motion associated with said subject's respiration are contained in a first few eigenvectors;
selecting one eigenvector from said first few eigenvectors; and
extracting a respiration rate for said subject from said selected eigenvector.
18. The system of claim 17 , wherein said feature vector comprises any of: an intensity component, a location component, and a motion component.
19. The system of claim 15 , wherein said video imaging device is a color video camera comprising red, green and blue channels and said intensity component is obtained from any of said channels.
20. The system of claim 15 , wherein said video imaging device is a color video camera comprising red, green and blue channels and said intensity component is obtained from computing an algebraic sum of pixel values from said red, green and blue channels.
21. The system of claim 20 , wherein, in advance of forming said feature vectors, grouping pixels associated with said features and computing their mean values.
22. The system of claim 21 , further comprising spatial filtering said grouped pixels to reduce noise.
23. The system of claim 21 , further comprising amplitude filtering by taking an L2 norm of each row vector and retaining all feature points with an L2 norm that is below a pre-defined maximum value.
24. The system of claim 17 , wherein said video matrix comprises orthonormal eigenvectors obtained by computing mean and covariance matrices using singular value decomposition.
25. The system of claim 24 , wherein, in advance of computing said mean and covariance matrices, computing a derivative of said video matrix.
26. The system of claim 17 , wherein eigenvectors are computed by performing singular value decomposition on said video matrix.
27. The system of claim 26 , wherein, in advance of performing singular value decomposition, computing a derivative of said video matrix.
28. The system of claim 17 , wherein extracting said respiration rate from said selected eigenvector comprises:
generating a power spectral density curve for said selected eigenvector; and
identifying a frequency within said power spectral density curve with a highest signal strength, said identified frequency being said subject's respiration rate.
29. The system of claim 17 , wherein extracting said respiration rate from said selected eigenvector comprises:
computing peak to peak interval between adjacent peaks; and
computing frequency from select group of peak to peak intervals.
30. The system of claim 17 , wherein, in response to said subject's respiration rate not being within a normal range, performing any of: initiating an alert, and signaling a medical professional.
31. The system of claim 17 , further comprising using said respiratory rate to facilitate a determination of an occurrence of any of: Sudden Infant Death Syndrome, Respiratory Distress, Respiratory Failure, and Pulmonary Disease.