Baby monitor
Systems and methods are described herein for monitoring and capturing information associated with target subjects, such as babies. For example, the systems and methods provide a capture device, such as a camera, that captures capture time of flight (TOF) data from a target subject. Using the TOF data, the systems and methods, via a display device associated with (e.g., paired to) the capture device, present information about the target subject, such as information indicating movement of the target subject, information indicating a breathing rate of the target subject, information indicating a heart rate of the target subject, and so on.
1. A method of detecting body movement status and body breathing status in a target subject, said method comprising the steps of:
receiving time of flight (TOF) data measured from said target subject;
parsing the TOF data to extract a phase frame and an amplitude frame from the TOF data;
extracting distance data for pixels from said phase frame, and extracting confidence data for said pixels from said amplitude frame;
determining pixel velocities for said pixels using said distance data of said phase frame;
masking said pixels to nullify any of said pixels below a selected amplitude threshold, therein producing high confidence pixels, wherein masking is performed using a mask generated from said confidence data of said amplitude frame;
filtering said high confidence pixels to produce filtered high confidence pixels;
thresholding said filtered high confidence pixels to identify areas of movement in a time series of frames;
performing blob analysis to identify blob centroids within said time series of frames;
determining spatial co-location by clustering said blob centroids that were identified across a said time series of frames;
determining statistics for said blob centroids that were identified;
aggregating said blob centroids that were identified into classes of clusters based on an analysis of said statistics, wherein said classes of clusters include a first class representative of said body movement status and a second class representative of said body breathing status;
extracting said body movement status and said body breathing status from said classes of clusters.
2. The method of claim 1 , wherein said TOF data received is captured by a TOF sensor proximate to the said target subject.
3. The method of claim 1 , wherein determining pixel velocities includes subtracting distance data extracted from a previous first phase frame from distance data extracted from a second phase frame.
4. The method of claim 1 , wherein filtering said high confidence pixels includes: applying a Gaussian low-pass filter to said high confidence pixels to spatially smooth said high confidence pixels; and performing temporal smoothing to select temporally correlated pixels of said high confidence pixels.
5. The method of claim 1 , further comprising: after thresholding said filtered high confidence pixels, performing morphological erosion to identify relatively larger areas of movement within said time series of frames.
6. The method of claim 1 , wherein performing blob analysis to identify blob centroids within said time series of frames includes:
performing blob analysis to identify centroids of morphologically connected regions between frames in said time series of frames; and
aggregating said centroids of morphologically connected regions identified across said time series of frames.
7. The method of claim 1 , wherein aggregating said blob centroids into classes of clusters based on an analysis of said statistics includes identifying said aggregated centroids using an iterated k-means algorithm.
8. The method of claim 1 , wherein extracting said body movement status and said body breathing status from said classes of clusters includes: processing using a fuzzy controller; and classifying truth values output by said fuzzy controller using an artificial neural network to determine said body movement status and said body breathing status.