Sleep tracking and vital sign monitoring using low power radio waves

1 . A system that performs radar-based measurement of vital signs, the system comprising:

a processing system, comprising one or more processors, that is in communication with a radar subsystem, the processing system being configured to:

analyze radar waveform data;

perform, using a trained machine learning arrangement, a classification based on multiple spectral characteristics of the analyzed radar waveform data to determine a state of a monitored region, wherein:

the trained machine learning arrangement is trained using a set of training data in which the set of training data is classified as mapping multiple spectral characteristics to ground-truth user states; and

the trained machine learning arrangement classifies the analyzed radar waveform data into a plurality of states, comprising a first state in which a user is present in the monitored region and the user is performing vitals-only movement and a second state which is different from the first state;

initiate, based on the state of the monitored region being in the first state, vital sign monitoring of the user;

determine, using the trained machine learning arrangement based on analyzing the radar waveform data, one or more vital signs of the user present within the monitored region; and

record the one or more determined vital signs of the user.

2 . The system of claim 1 , wherein the system further comprises:

a housing;

an RF emitter located within the housing; and

an RF receiver located within the housing, wherein the housing is configured to be positioned such that a field of view of the RF receiver is pointed toward the monitored region that includes the user in bed.

3 . The system of claim 2 , further comprising a radar processing circuit located within the housing that processes data received from the RF receiver and outputs raw waveform data.

4 . The system of claim 3 , wherein the processing system is located within the housing.

5 . The system of claim 3 , wherein the processing system is further configured to:

receive the raw waveform data from the radar processing circuit, wherein the processing system being configured to analyze the radar waveform data comprises the processing system being configured to:

filter, from the raw waveform data, waveform data indicative of static objects to obtain motion-indicative waveform data.

6 . The system of claim 5 , wherein the processing system is further configured to:

filter one or more frequencies of the filtered waveform data to sum harmonics of the one or more frequencies caused by the one or more vital signs.

7 . The system of claim 2 , wherein the RF emitter emits frequency-modulated continuous-wave (FMCW) radar.

8 . The system of claim 1 , wherein the second state indicates the user is present and moving.

9 . The system of claim 1 , wherein the second state indicates the user is not present.

10 . The system of claim 1 , wherein the processing system being configured to perform, using the trained machine learning arrangement, the classification comprises the processing system being further configured to use a neural network to perform the classification based on spectral energy and spectral sparsity.

11 . The system of claim 1 , further comprising:

a smartphone housing the processing system; and

a smartphone charger stand, wherein the smartphone is docked with the smartphone charger stand such that a radar sensor of the smartphone is pointed to emit radio waves at a location where the user will be sleeping.

12 . The system of claim 1 , wherein the one or more vital signs comprises a breathing rate, a heartrate, or both.

13 . A method for radar-based measurement of vital signs, the method comprising:

analyzing by a processing system comprising one or more processors, radar waveform data;

performing, by the processing system, using a trained machine learning arrangement, a classification based on multiple spectral characteristics of the analyzed radar waveform data to determine a state of a monitored region, wherein:

the trained machine learning arrangement is trained using a set of training data in which the set of training data is classified as mapping multiple spectral characteristics to ground-truth user states; and

the trained machine learning arrangement classifies the analyzed radar waveform data into a plurality of states, comprising a first state in which a user is present and is performing vitals-only movement and a second state which is different from the first state;

permitting, by the processing system based on the monitored region being in the first state, vital sign monitoring;

determining, by the processing system using the trained machine learning arrangement based on analyzing the radar waveform data, one or more vital signs of the user present within the monitored region; and

recording, by the processing system to a non-transitory processor-readable medium, an indication of the one or more determined vital signs of the user.

14 . The method of claim 13 , wherein performing the classification comprises using, by the processing system, a neural network to perform the classification based on spectral energy and spectral sparsity.

15 . The method of claim 13 , further comprising:

emitting, by an RF emitter located within a same housing as the processing system, radio waves, wherein the emitted radio waves are emitted as part of frequency-modulated continuous-wave (FMCW) radar waves; and

receiving, by an RF receiver located within the same housing as the processing system, reflections of the radio waves.

16 . The method of claim 15 , further comprising: processing, by a radar processing circuit distinct from the processing system, the reflections of the radio waves from the RF receiver and outputs raw waveform data.

17 . The method of claim 16 , further comprising: filtering, by the processing system, from the raw waveform data, waveform data indicative of static objects to obtain motion-indicative waveform data.

18 . The method of claim 16 , wherein the processing system, the radar processing circuit, the RF emitter, and the RF receiver are part of a smartphone, the method further comprising:

docking the smartphone with a charger stand such that the RF emitter of the smartphone is pointed to emit the radio waves at the monitored region where the user will be sleeping.

19 . A non-transitory processor-readable medium comprising process-readable instructions configured to cause one or more processors to:

analyze radar waveform data;

perform, using a trained machine learning arrangement, a classification based on multiple spectral characteristics of the analyzed radar waveform data to determine a state of a monitored region, wherein:

the trained machine learning arrangement is trained using a set of training data in which the set of training data is classified as mapping multiple spectral characteristics to ground-truth user states; and

the trained machine learning arrangement classifies the analyzed radar waveform data into a plurality of states, comprising a first state in which a user is present in the monitored region and is performing vitals-only movement and a second state which is different than the first state;

initiate, based on the state of the monitored region being in the first state, vital sign monitoring of the user;

determine, using the trained machine learning arrangement based on the analyzing the radar waveform data, one or more vital signs of the user present within the monitored region; and

record the one or more determined vital signs of the user.

20 . The non-transitory processor-readable medium of claim 19 , wherein the one or more processors are configured to use a neural network to perform the classification based on spectral energy and spectral sparsity.