Granted Patent B2

US 12,474,781 · App. 17/821,732 · Granted Nov 18, 2025

Inventors: Wenxun Qiu (Allen, TX); Boon Loong Ng (Plano, TX); Han Wang (Plano, TX); Khuong N. Nguyen (Allen, TX)

Assignee: Samsung Electronics Co., Ltd.

G06F3/017G01S7/415G01S13/581G01S13/88G06F3/046

Abstract

Claims (114)

1 . An electronic device for gesture recognition, the electronic device comprising:

a radar transceiver; and

a processor operably connected to the radar transceiver, the processor configured to:

transmit, via the radar transceiver, radar signals;

extract features from reflections of the radar signals received by the radar transceiver;

determine whether an activity is detected based on the features;

in response to a determination that the activity is detected, classify a gesture as one from among N different gestures in a dataset and an invalid gesture from the features, wherein:

each of the N different gestures is categorized into one of two categories that include a signature gesture and a non-signature gesture;

the N different gestures in the dataset comprise a plurality of signature gestures and a plurality of non-signature gestures;

the signature gesture is defined by at least one repetition in one or more feature domains, the at least one repetition including a repeat, a flipped version, or a movement in an opposite direction of another movement;

the non-signature gesture is defined by no repetition in the one or more feature domains;

when classified from among the plurality of signature gestures, the classified gesture is thereby the signature gesture; and

when classified from among the plurality of non-signature gestures, the classified gesture is thereby the non-signature gesture;

compare a set of ranges associated with the features to a set of thresholds associated with the classified gesture to verify the classified gesture;

determine whether the classified gesture corresponds to an action for the electronic device to perform, when the classified gesture is the signature gesture;

generate an output to perform the action corresponding to the classified gesture that was verified based on a result of the comparison, in response to a determination that the signature gesture classified corresponds to the action; and

activate a function scenario, in response to a determination that the signature gesture classified does not correspond to the action.

2 . The electronic device of claim 1 , wherein the processor is further configured to:

identify differences between the features over a period of time; and

determine that the activity is detected based on differences between the features.

3 . The electronic device of claim 2 , wherein the processor is further configured to:

identify a region of interest corresponding to the differences between the features over the period of time, and

classify the gesture using a portion of the radar signals that correspond to the region of interest.

4 . The electronic device of claim 1 , wherein the processor is further configured to:

store the features extracted from the reflections in a buffer; and

reset the buffer after the output is generated.

5 . The electronic device of claim 1 , wherein:

the features include a time-velocity diagram (TVD) indicating a radial velocity between the radar transceiver and an object that is moving; and

the processor is further configured to:

identify the TVD based on a maximum power, and

classify the gesture using the TVD.

6 . The electronic device of claim 1 , wherein the electronic device further comprises a neural network configured to classify the gesture; and

wherein the action is to at least one of:

activate the function scenario that enables the electronic device to perform an action corresponding to a non-signature gesture classified during the activated function scenario; or

provide a notification to a user of the electronic device that informs the user to perform a non-signature gesture.

7 . The electronic device of claim 1 , wherein:

the set of ranges include at least one of: a movement range, a Doppler shift range, an azimuth angle range, an elevation angle range, and a power level range; and

the processor is further configured to:

identify the set of ranges from the features, and

select the set of thresholds based on the classified gesture.

8 . The electronic device of claim 7 , wherein:

to identify the movement range, the processor is configured to identify a difference between a furthest distance between an object and the radar transceiver and a closest distance between the object and the radar transceiver during a predefined time;

to identify the Doppler shift range, the processor is configured to identify differences between Doppler components during the predefined time;

to identify the azimuth angle range, the processor is configured to identify a difference between a first azimuth angle detected during the predefined time and a second azimuth angle detected during the predefined time, the first azimuth angle representing a maximum angle in azimuth and the second azimuth angle representing a minimum angle in azimuth;

to identify the elevation angle range, the processor is configured to identify a difference between a first elevation angle detected during the predefined time and a second elevation angle detected during the predefined time, the first elevation angle representing a maximum angle in elevation and the second elevation angle representing a minimum angle in elevation; and

to identify the power level range, the processor is configured to identify a change in power during the predefined time.

9 . The electronic device of claim 7 , wherein to generate the result of the comparison, the processor is configured to:

select a four thresholds that correspond a drumming gesture, when the gesture is classified as the drumming gesture;

determine that the movement range is less than a first of the four thresholds;

determine that the Doppler shift range is larger than a second of the four thresholds;

determine that the azimuth angle range is less than a third of the four thresholds; and

determine that the power level range is larger than a fourth of the four thresholds.

10 . The electronic device of claim 7 , wherein to generate the result of the comparison, the processor is configured to:

select a four thresholds that correspond a waving gesture, when the gesture is classified as the waving gesture;

determine that the movement range is larger than a first of the four thresholds;

determine that the Doppler shift range is larger than a second of the four thresholds;

determine that the azimuth angle range is larger than a third of the four thresholds; and

determine that the elevation angle range is less than a fourth of the four thresholds.

11 . A method for gesture recognition, the method comprising:

transmitting, via a radar transceiver, radar signals;

extracting features from reflections of the radar signals received by the radar transceiver;

determining whether an activity is detected based on the features;

in response to a determination that the activity is detected, classifying a gesture as one from among N different gestures in a dataset and an invalid gesture from the features, wherein:

each of the N different gestures is categorized into one of two categories that include a signature gesture and a non-signature gesture;

the N different gestures in the dataset comprise a plurality of signature gestures and a plurality of non-signature gestures;

the signature gesture is defined by at least one repetition in one or more feature domains, the at least one repetition including a repeat, a flipped version, or a movement in an opposite direction of another movement;

the non-signature gesture is defined by no repetition in the one or more feature domains;

when classified from among the plurality of signature gestures, the classified gesture is thereby the signature gesture; and

when classified from among the plurality of non-signature gestures, the classified gesture is thereby the non-signature gesture;

comparing a set of ranges associated with the features to a set of thresholds associated with the classified gesture to verify the classified gesture;

determining whether the classified gesture corresponds to an action for an electronic device to perform, when the classified gesture is the signature gesture;

generating an output to perform the action corresponding to the classified gesture that was verified based on a result of the comparison, in response to a determination that the signature gesture classified corresponds to the action; and

activating a function scenario, in response to a determination that the signature gesture classified does not correspond to the action.

12 . The method of claim 11 , further comprising:

identifying differences between the features over a period of time; and

determining that the activity is detected based on differences between the features.

13 . The method of claim 12 , further comprising:

identifying a region of interest corresponding to the differences between the features over the period of time, and

classifying the gesture using a portion of the radar signals that correspond to the region of interest.

14 . The method of claim 11 , further comprising:

storing the features extracted from the reflections in a buffer; and

resetting the buffer after the output is generated.

15 . The method of claim 11 , wherein:

the features include a time-velocity diagram (TVD) indicating a radial velocity between the radar transceiver and an object that is moving; and

the method further comprises:

identifying the TVD based on a maximum power, and

classifying the gesture using the TVD.

16 . The method of claim 11 , wherein a neural network classifies the gesture; and

wherein the action is to at least one of:

activate the function scenario that enables the electronic device to perform an action corresponding to a non-signature gesture classified during the activated function scenario; or

provide a notification to a user of the electronic device that informs the user to perform a non-signature gesture.

17 . The method of claim 11 , wherein:

the set of ranges include at least one of: a movement range, a Doppler shift range, an azimuth angle range, an elevation angle range, and a power level range; and

the method further comprises:

identifying the set of ranges from the features, and

selecting the set of thresholds based on the classified gesture.

18 . The method of claim 17 , wherein:

identifying the movement range comprises identifying a difference between a furthest distance between an object and the radar transceiver and a closest distance between the object and the radar transceiver during a predefined time;

identifying the Doppler shift range comprises identifying differences between Doppler components during the predefined time;

identifying the azimuth angle range comprises identifying a difference between a first azimuth angle detected during the predefined time and a second azimuth angle detected during the predefined time, the first azimuth angle representing a maximum angle in azimuth and the second azimuth angle representing a minimum angle in azimuth;

identifying the elevation angle range comprises identifying a difference between a first elevation angle detected during the predefined time and a second elevation angle detected during the predefined time, the first elevation angle representing a maximum angle in elevation and the second elevation angle representing a minimum angle in elevation; and

identify the power level range comprises identifying a change in power during the predefined time.

19 . The method of claim 17 , wherein generating the result of the comparison, comprises:

selecting a four thresholds that correspond a drumming gesture, when the gesture is classified as the drumming gesture;

determining that the movement range is less than a first of the four thresholds;

determining that the Doppler shift range is larger than a second of the four thresholds;

determining that the azimuth angle range is less than a third of the four thresholds; and

determining that the power level range is larger than a fourth of the four thresholds.

20 . The method of claim 17 , wherein generating the result of the comparison, comprises:

selecting a four thresholds that correspond a waving gesture, when the gesture is classified as the waving gesture;

determining that the movement range is larger than a first of the four thresholds;

determining that the Doppler shift range is larger than a second of the four thresholds;

determining that the azimuth angle range is larger than a third of the four thresholds; and

determining that the elevation angle range is less than a fourth of the four thresholds.

Assignments (1)

Continuity (4)

References Cited (61)