System and method for determining engagement level of a human being using a millimeter-wave radar sensor
An embodiment method may include receiving radar data at a millimeter-wave radar sensor, the radar data being generated in response to an incident radio-frequency signal reflecting off a biological target located in a field of view of the millimeter-wave radar sensor; extracting a filtered vital-Doppler signal from the radar data; determining an emotion of the biological target based on the filtered vital-Doppler signal; and generating a user-engagement rating indicative of a level of engagement of the biological target in an activity performed in the field of view based on the emotion.
1. A method, comprising:
receiving radar data at a millimeter-wave radar sensor, the radar data being generated in response to an incident radio-frequency signal reflecting off a biological target located in a field of view of the millimeter-wave radar sensor;
extracting a filtered vital-Doppler signal from the radar data, wherein extracting the filtered vital-Doppler signal from the radar data comprises:
capturing the radar data corresponding to the biological target across a vital-Doppler frame;
performing a first filtering on the vital-Doppler frame to produce a first-filtered vital-Doppler frame; and
performing a second filtering on the vital-Doppler frame to produce a second-filtered vital-Doppler frame;
designating at least one of the first-filtered vital-Doppler frame or the second-filtered vital-Doppler frame as a time-domain vital-Doppler signal;
filtering the time-domain vital-Doppler signal with a window function that is localized in time and frequency to generate a windowed time-frequency signal, the filtering the time-domain vital-Doppler signal comprising convolving the time-domain vital-Doppler signal with the window function;
determining initial filter coefficients of a frequency filter;
performing a Gabor time-frequency transform on the windowed time-frequency signal;
filtering a result of the Gabor time-frequency transform step using the frequency filter having the initial filter coefficients;
performing an inverse Gabor time-frequency transform on an output of the frequency filter to generate an interim filtered vital-Doppler signal;
determining a ratio of an out-of-band energy of the interim filtered vital-Doppler signal to a total energy of the interim filtered vital-Doppler signal; and
in response to the ratio being greater than a threshold, iteratively performing the following steps:
selecting new filter coefficients of the frequency filter based on a predetermined sequence of filter coefficients;
performing the Gabor time-frequency transform;
filtering the result of the Gabor time-frequency transform step using the frequency filter having the new filter coefficient; and
performing the inverse Gabor time-frequency transform;
determining an emotion of the biological target based on the filtered vital-Doppler signal; and
generating a user-engagement rating indicative of a level of engagement of the biological target in an activity performed in the field of view based on the emotion.
2. The method of claim 1 , wherein determining the emotion of the biological target based on the filtered vital-Doppler signal comprises:
determining a number of times a respective emotion of a plurality of emotions occurred during the activity.
3. The method of claim 2 , wherein generating the user-engagement rating comprises:
generating the user-engagement rating based on the plurality of emotions and the number of times each of the plurality of emotions occurred during the activity.
4. The method of claim 1 , wherein determining the emotion of the biological target based on the filtered vital-Doppler signal comprises:
determining a time instance at which a respective emotion of a plurality of emotions occurred during the activity; and
determining a probability of the respective emotion occurring at the time instance.
5. The method of claim 4 , wherein generating the user-engagement rating comprises:
generating the user-engagement rating based on the plurality of emotions, the time instance associated with each of the plurality of emotions, and the probability associated with each of the plurality of emotions.
6. The method of claim 1 , wherein determining the emotion of the biological target based on the filtered vital-Doppler signal comprises determining respective emotions of each of a plurality of biological targets based on the filtered vital-Doppler signal.
7. The method of claim 6 , wherein generating the user-engagement rating indicative of the level of engagement of the biological target in the activity based on the emotion comprises generating a respective user-engagement rating indicative of a respective level of engagement of each of the plurality of biological targets based on the respective emotions of each of the plurality of biological targets.
8. The method of claim 7 , further comprising generating a collective statistic indicative of the level of engagement of the plurality of biological targets based on the respective user-engagement ratings.
9. The method of claim 1 , wherein the first filtering comprises a band-pass filtering centered at 0.8 Hz with a bandwidth of 0.6 Hz.
10. The method of claim 1 , wherein the second filtering comprises a band-pass filtering centered at 2 Hz with a bandwidth of 3 Hz.
11. The method of claim 1 , wherein a result of the inverse Gabor time-frequency transform is designated as the filtered vital-Doppler signal in response to the ratio being less than the threshold.
12. The method of claim 1 , wherein the out-of-band energy of the interim filtered vital-Doppler signal comprises an energy of the interim filtered vital-Doppler signal outside a frequency range of 0.6 Hz to 5 Hz.
13. The method of claim 1 , wherein determining the emotion of the biological target based on the filtered vital-Doppler signal comprises:
performing a set of reference radar measurements for a plurality of reference emotion labels on the biological target using the millimeter-wave radar sensor;
producing a training data set for the plurality of reference emotion labels based on the set of reference radar measurements;
forming a set of stored feature vectors based on the training data set; and
associating a respective emotion label to each stored feature vector of the set of stored feature vectors.
14. The method of claim 13 , wherein forming the set of stored feature vectors comprises iteratively adjusting the forming of the set stored feature vectors using a machine learning algorithm.
15. The method of claim 14 , wherein determining the emotion of the biological target based on the filtered vital-Doppler signal comprises:
extracting live feature vectors from the filtered vital-Doppler signal; and
determining the emotion of the biological target based on the live feature vectors and the set stored feature vectors using the machine learning algorithm.
16. A system, comprising:
a processing system configured to be coupled to a millimeter-wave radar sensor, wherein the processing system is configured to:
instruct the millimeter-wave radar sensor to transmit a series of chirps within a field of view of the millimeter-wave radar sensor;
identify a set of targets within the field of view based on radar data received by the millimeter-wave radar sensor and in response to transmission of the series of chirps;
extract a filtered vital-Doppler signal from the radar data, wherein extracting the filtered vital-Doppler signal from the radar data comprises:
capturing the radar data corresponding to the set of targets across a vital-Doppler frame;
performing a first filtering on the vital-Doppler frame to produce a first-filtered vital-Doppler frame; and
performing a second filtering on the vital-Doppler frame to produce a second-filtered vital-Doppler frame;
designating at least one of the first-filtered vital-Doppler frame or the second-filtered vital-Doppler frame as a time-domain vital-Doppler signal;
filtering the time-domain vital-Doppler signal with a window function that is localized in time and frequency to generate a windowed time-frequency signal, the filtering the time-domain vital-Doppler signal comprising convolving the time-domain vital-Doppler signal with the window function;
determining initial filter coefficients of a frequency filter;
performing a Gabor time-frequency transform on the windowed time-frequency signal;
filtering a result of the Gabor time-frequency transform step using the frequency filter having the initial filter coefficients;
performing an inverse Gabor time-frequency transform on an output of the frequency filter to generate an interim filtered vital-Doppler signal;
determining a ratio of an out-of-band energy of the interim filtered vital-Doppler signal to a total energy of the interim filtered vital-Doppler signal; and
in response to the ratio being greater than a threshold, iteratively performing the following steps:
selecting new filter coefficients of the frequency filter based on a predetermined sequence of filter coefficients;
performing the Gabor time-frequency transform;
filtering the result of the Gabor time-frequency transform step using the frequency filter having the new filter coefficient; and
performing the inverse Gabor time-frequency transform;
determine an emotion of the set of targets based on the filtered vital-Doppler signal; and
generate a user-engagement rating indicative of a level of engagement of the set of targets in an activity performed in the field of view based on the emotion.
17. A non-transitory computer readable storage medium with an executable program stored thereon, the executable program including instructions to:
instruct a millimeter-wave radar sensor to transmit a series of chirps within a field of view of the millimeter-wave radar sensor;
identify a biological target within the field of view based on radar data received by the millimeter-wave radar sensor and in response to transmission of the series of chirps;
extract a filtered vital-Doppler signal from the radar data, wherein extracting the filtered vital-Doppler signal from the radar data comprises:
capturing the radar data corresponding to the biological target across a vital-Doppler frame;
performing a first filtering on the vital-Doppler frame to produce a first-filtered vital-Doppler frame; and
performing a second filtering on the vital-Doppler frame to produce a second-filtered vital-Doppler frame;
designating at least one of the first-filtered vital-Doppler frame or the second-filtered vital-Doppler frame as a time-domain vital-Doppler signal;
filtering the time-domain vital-Doppler signal with a window function that is localized in time and frequency to generate a windowed time-frequency signal, the filtering the time-domain vital-Doppler signal comprising convolving the time-domain vital-Doppler signal with the window function;
determining initial filter coefficients of a frequency filter;
performing a Gabor time-frequency transform on the windowed time-frequency signal;
filtering a result of the Gabor time-frequency transform step using the frequency filter having the initial filter coefficients;
performing an inverse Gabor time-frequency transform on an output of the frequency filter to generate an interim filtered vital-Doppler signal;
determining a ratio of an out-of-band energy of the interim filtered vital-Doppler signal to a total energy of the interim filtered vital-Doppler signal; and
in response to the ratio being greater than a threshold, iteratively performing the following steps:
selecting new filter coefficients of the frequency filter based on a predetermined sequence of filter coefficients;
performing the Gabor time-frequency transform;
filtering the result of the Gabor time-frequency transform step using the frequency filter having the new filter coefficient; and
performing the inverse Gabor time-frequency transform;
determine an emotion of the biological target based on the filtered vital-Doppler signal; and
generate a user-engagement rating indicative of a level of engagement of the biological target in an activity performed in the field of view based on the emotion.