Knuckle operation identification method and electronic device

Patent metadata

Field	Value
Publication number	US-12422955-B2
Application number	US-202318840247-A
Country	US
Kind code	B2
Filing date	Apr 28, 2023
Priority date	Jul 1, 2022
Publication date	Sep 23, 2025
Grant date	Sep 23, 2025

Abstract

This application relates to the field of artificial intelligence technologies, and provide a knuckle operation identification method and an electronic device. According to this application, when a knuckle performs an operation on different touch regions of a touch panel, an acceleration signal may have different change trends. A touch feature is extracted from a touch signal for determining a contact area and a contact location, and an acceleration feature is extracted from the acceleration signal for determining a magnitude of a touch on a screen. A capacitor binary classification model is used at the front, to extract a confidence score indicating a correlation to a knuckle from a capacitor signal. Then, feature fusion is performed on the acceleration feature, the confidence score, and the touch feature, and a fused feature is input into a knuckle classification model, to obtain a better classification effect.

First claim

The invention claimed is:

1. A knuckle operation identification method, wherein the method comprises:
obtaining an acceleration signal, a capacitor signal, and a touch signal in response to a touch operation on a touch panel, wherein the acceleration signal is a raw signal acquired by an acceleration sensor, the capacitor signal is a raw signal acquired by a capacitance sensor, and the touch signal is a signal obtained by processing the capacitor signal;
extracting an acceleration feature, a confidence score, and a touch feature, wherein the acceleration feature is a feature, in the acceleration signal, associated with a knuckle operation, the confidence score is a feature, in the capacitor signal, associated with the knuckle operation, and the touch feature is a feature, in the touch signal, associated with the knuckle operation;
performing feature fusion on the acceleration feature, the confidence score, and the touch feature to obtain a fused feature; and
inputting the fused feature into a knuckle classification model to obtain a target classification result, wherein the target classification result indicates whether the touch operation is a knuckle operation or a non-knuckle operation.

2. The method according to claim 1 , wherein the confidence score is a feature extracted from the touch signal based on the acceleration feature indicating that the touch operation is a knuckle operation.

3. The method according to claim 2 , wherein the extracting the acceleration feature, the confidence score, and the touch feature comprises:
extracting the acceleration feature from the acceleration signal;
inputting the acceleration feature into an acceleration binary classification model to obtain a preliminary classification result; and
based on the preliminary classification result indicating that the touch operation is a knuckle operation, inputting the capacitor signal into a capacitance binary classification model to obtain the confidence score, and extracting the touch feature from the touch signal.

4. The method according to claim 3 , wherein the capacitance binary classification model is a convolutional neural network model, and the acceleration binary classification model is a fully connected neural network model.

5. The method according to claim 2 , wherein the extracting the acceleration feature, the confidence score, and the touch feature comprises:
extracting the acceleration feature from the acceleration signal, and extracting the touch feature from the touch signal;
inputting the acceleration feature into an acceleration binary classification model to obtain a preliminary classification result; and
based on the preliminary classification result indicating that the touch operation is a knuckle operation, inputting the capacitor signal into a capacitance binary classification model to obtain the confidence score.

6. The method according to claim 1 ,
wherein the touch feature comprises at least one of a contact location feature or a contact area feature; and
wherein the contact location feature indicates a location of interaction between a body part and the touch panel, and the contact area feature indicates a contact area of the body part with the touch panel.

7. The method according to claim 6 ,
wherein the contact location feature indicates a square coordinate number of a square in which a touch point is located; and
wherein the square in which the touch point is located is at least one square in a grid obtained by dividing the touch panel based on resolution of the touch panel.

8. The method according to claim 7 , wherein the grid comprises p rows and q columns of squares, a length of each square in the grid is equal to a quantity of pixels on a vertical axis of the touch panel divided by p, and a width of each square in the grid is equal to a quantity of pixels on a horizontal axis of the touch panel divided by q, wherein p and q are positive integers.

9. The method according to claim 7 ,
wherein the touch feature comprises the contact location feature; and
wherein extracting the contact location feature comprises:
determining an x coordinate and a y coordinate of the touch point based on the touch signal; and
determining, based on the x coordinate and the y coordinate, the contact location feature indicating the square coordinate number of the square in which the touch point is located, wherein an x-axis is a horizontal direction of a plane on which the touch panel is located, and a y-axis is a vertical direction of the plane on which the touch panel is located.

10. The method according to claim 1 , wherein the acceleration feature comprises at least one of the following: a maximum gradient feature, a signal amplitude feature, a front-part zero cross counting feature, a maximum high pass feature, a mean-add-absolute-value feature, a front-part normalized value square error feature, a front-part normalized value amplitude feature, a fast Fourier transform mean feature, or a part fast Fourier transform mean feature.

11. The method according to claim 1 , wherein the confidence score is a score on a degree of confidence, and the score on the degree of confidence indicates a degree of association between the capacitor signal and a knuckle operation.

12. The method according to claim 1 , the knuckle classification model is a fully connected neural network model.

13. The method according to claim 1 , wherein the method further comprises:
determining, based on the target classification result indicating that the touch operation is a knuckle operation, a knuckle gesture to which the touch operation belongs, and performing a responsive function corresponding to the knuckle gesture, wherein different knuckle gestures correspond to different responsive functions.

14. The method according to claim 13 , wherein the knuckle gesture comprises at least one of the following: a gesture of knocking twice with a knuckle, a gesture of knocking and drawing an enclosed region with a knuckle, a gesture of knocking with a knuckle and drawing a letter S, a gesture of swiping down on a screen from top to bottom with three knuckles, a gesture of knocking twice with two knuckles, or a gesture of knocking and drawing a straight line in the middle of the screen with a knuckle; wherein a first responsive function corresponding to the gesture of knocking twice with a knuckle is a function of taking a full screenshot; wherein a second responsive function corresponding to the gesture of knocking and drawing an enclosed region with a knuckle is a function of capturing part of the screen; wherein a third responsive function corresponding to the gesture of knocking with a knuckle and drawing a letter S is a function of taking a scroll shot; wherein a fourth responsive function corresponding to the gesture of swiping down on a screen from top to bottom with three knuckles is a function of swiping to take a screenshot; wherein a fifth responsive function corresponding to the gesture of knocking twice with two knuckles is a function of starting or ending screen recording; and wherein a sixth responsive function corresponding to the gesture of knocking and drawing a straight line in the middle of the screen with a knuckle is a function of screen splitting.

15. An electronic device, comprising:
a memory storing a computer program or instructions; and
a processor, wherein the processor is coupled to the memory, and the processor is configured to execute the computer program or instructions stored in the memory, to enable the electronic device to implement a knuckle operation identification method comprising:
obtaining an acceleration si

Assignees

• Honor Device Co Ltd

Inventors

• Wang Xiaochen
• Zhang Cheng
• Wan Nuliang

Classifications

• G06F3/0446

  using a grid-like structure of electrodes in at least two directions, e.g. using row and column electrodes · CPC title

• G06V10/80

  Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level (multimodal speaker identification or verification G10L17/10) · CPC title

• G06V40/1306

  non-optical, e.g. ultrasonic or capacitive sensing · CPC title

• G06F3/04186 Primary

  Touch location disambiguation · CPC title

• G06F3/044

  by capacitive means · CPC title