Fingerprint-based authentication using touch input

Abstract

Techniques and apparatus are described that enable a device to be unlocked and continuous user authentication to be enabled without a touch input specifically for a fingerprint request. Touch input (106) including one or more touches (108) on a touchscreen (306) is received, and raw image data (112) corresponding to the touch (108) is retrieved from a fingerprint imaging sensor (308). A pixel clustering technique is performed on the raw image data (112) to determine a portion of the raw image data (118) corresponding to each touch (108). A touch embedding (120) is formed for each portion of the raw image data (118), and the touch embedding is compared with one or more stored fingerprint embeddings (220) corresponding to the corresponding fingerprints of one or more authorized users. An authentication result (204) is then determined for the touch input (106) based on the comparison result (124).

Description

Background Technology

[0001] User authentication is a crucial tool for ensuring the security of sensitive information accessible on computing devices. Fingerprint sensors are often implemented as a fast and efficient user authentication device. Recently, under-display fingerprint sensors have been implemented, enabling fingerprint authentication in a user-friendly and space-saving manner.

[0002] However, in many situations, dedicated fingerprint input is required to authenticate users. For example, a device may require users to place their finger in a certain position or direction to unlock it. Since the input is only used to unlock the device, this can degrade the user experience. Furthermore, once the device is unlocked, unauthorized users can use it without hindrance. Summary of the Invention

[0003] Technologies and devices enabling fingerprint-based authentication using touch input are described. These technologies and devices allow computing devices to be unlocked and allow users to be continuously authenticated without explicit fingerprint touch input. This reduces the amount of touch input required for a user to unlock the computing device and improves security by allowing continuous authentication. For example, when the device is locked, a soft lock screen (e.g., a blurred user interface or an overlay on top of the user interface) can be presented. By simply interacting with the soft-locked user screen (e.g., through touch input to user interface elements), the device can unlock, remove the soft lock screen, or otherwise change to the underlying user interface, perform actions corresponding to the touch input, and allow the user to continue interacting with the device as usual. Once unlocked, the device can periodically authenticate touch input to ensure that the touch input still originates from an authorized user. Touch input can be associated with specific actions (e.g., in addition to being specifically limited to fingerprint recognition), allowing the user to be authenticated and input actions within a single touch input.

[0004] The aspects described below include a method for fingerprint-based authentication performed by a computing device. The method includes receiving touch input, including one or more touches on a touchscreen of the computing device, and retrieving raw image data corresponding to the touches from a fingerprint imaging sensor of the computing device. The method also involves performing pixel clustering techniques on the raw image data to determine portions of the raw image data corresponding to each touch, and establishing one or more comparison results for the touch input. The comparison results are established for each touch by forming a touch embedding representing a portion of the raw image data corresponding to the respective touch, calculating the similarity between the touch embedding and each of one or more stored fingerprint embeddings corresponding to the fingerprints of one or more authorized users, and establishing a comparison result for the respective touch based on the calculation. The method then involves determining an authentication result for the touch input based on the comparison results.

[0005] The aspects described below also include a computing device comprising a touchscreen, a fingerprint imaging sensor, at least one processor, and at least one computer-readable storage medium including instructions that, when executed by the processor, cause the processor to receive touch input including one or more touches on the touchscreen and retrieve raw image data corresponding to the touches from the fingerprint imaging sensor. The instructions also cause the processor to perform pixel clustering on the raw image data to determine portions of the raw image data corresponding to each touch and to establish one or more comparison results for the touch input. The comparison results are established for each touch by forming a touch embedding representing a portion of the raw image data corresponding to the corresponding touch, calculating the similarity between the touch embedding and each of one or more stored fingerprint embeddings corresponding to the fingerprints of one or more authorized users, and establishing a comparison result for the corresponding touch based on the calculation. The instructions then cause the processor to determine the authentication result of the touch input based on the comparison results. Attached Figure Description

[0006] The following figures illustrate devices and techniques that enable fingerprint-based authentication using touch input. The same numbers are used throughout the figures to refer to similar features and components:

[0007] Figure 1 The diagram illustrates an example process flow for fingerprint-based authentication using touch input;

[0008] Figure 2 The illustration shows a sample data stream for a fingerprint module used for fingerprint-based authentication using touch input;

[0009] Figure 3 The illustration shows an example computing device that can implement fingerprint-based authentication using touch input;

[0010] Figure 4 This is an example illustration of fingerprint-based authentication combined with a soft lock screen; and

[0011] Figure 5 The illustration shows an example method for fingerprint-based authentication using touch input. Detailed Implementation

[0012] Overview

[0013] Under-display fingerprint sensors enable fingerprint-based authentication in the absence of a dedicated fingerprint input area (e.g., a rear fingerprint sensor or a fingerprint sensor on a physical button). While under-display fingerprint sensors are more space-saving and intuitive for users, they typically require dedicated touch input for user authentication. Therefore, separate steps are usually required to interact with a locked device (e.g., Step 1: Unlock the device, Step 2: Interact with the device). Furthermore, once the device is unlocked, it may not be able to continuously authenticate the user during interaction (e.g., because typical interactions may not correspond to the location or orientation of dedicated touch input). Even if the device has been unlocked by an authorized user, this can lead to unauthorized access / use of the device.

[0014] The methods and apparatus described herein enable device unlocking and continuous user authentication without requiring dedicated touch input for fingerprint requests. For example, a soft lock screen can be presented when the device is locked. The device can be unlocked and functions (e.g., functions associated with user interface elements) can be performed by simply interacting with the underlying user interface (e.g., via touch input to user interface elements). Once the device is unlocked, touch input can be periodically authenticated as the user interacts with it. In this way, the transition from a locked to an unlocked state can be seamless, and the user can be continuously authenticated via normal interaction with the device.

[0015] Example process flow

[0016] Figure 1 The illustration shows an example process flow 100 for fingerprint-based authentication using touch input. Process flow 100 is typically implemented in computing device 102. Figure 3 The computing device 102 will be discussed further. Process flow 100 can begin at 104 with the computing device 102 in a locked or unlocked state. In either case, process flow 100 is generally the same. However, as described below, some details depend on the initial conditions.

[0017] At 104, the user interface is displayed by the touchscreen of the computing device, and touch input 106, including one or more touches 108, is received by the touchscreen. If the computing device 102 is in a locked state, the touch input 106 can be received by a soft lock screen (not shown) in conjunction with the user interface, such as regarding... Figure 4Further discussion. Touch input 106 may include a single touch (e.g., a single-finger tap or drag gesture) or multiple touches (e.g., multiple-finger tap or drag gestures or multi-finger pinch or spread gestures). In some cases, one or more of the touches 108 may correspond to something other than a finger (e.g., an object, a palm, a knuckle). Alternatively or additionally, touches 108 may originate from two or more users (e.g., all users are touching the touchscreen). In the example shown, touch input 106 includes a two-finger spread gesture with touches 108-1 and 108-2 (e.g., to initiate zoom).

[0018] Based on the received touch input 106, at 110, computing device 102 receives raw image data 112 from the fingerprint sensor under the display of computing device 102. The raw image data 112 corresponds to touch input 106.

[0019] At 114, a clustering technique is performed on the original image data 112. The clustering technique clusters the pixels of the original image data 112 into one or more clusters 116 (e.g., clusters 116-1 and 116-2), such that each of the clusters 116 corresponds to one of the touches 108. In the example shown, cluster 116-1 corresponds to touch 108-1, and cluster 116-2 corresponds to touch 108-2.

[0020] A portion of the original image data 118 corresponding to cluster 116 (e.g., a portion of original image data 118-1 and a portion of original image data 118-2) is individually passed through a neural network to determine touch embeddings 120 for each portion of the original image data 118 (e.g., touch embedding 120-1 for a portion of original image data 118-1 and touch embedding 120-2-2 for a portion of original image data 118-1). For example, the neural network may include radial convolutional layers to enforce rotation invariance (along with other suitable layers), and the touch embeddings 120 may include hashes of portions of the original image data 118.

[0021] At 122, computing device 102 can compare each touch embedding 120 with a stored fingerprint embedding of an authorized user to determine whether the corresponding touch embedding 120 matches or is sufficiently close to one of the stored fingerprint embeddings (e.g., within the radius of the embedding space). The stored fingerprint embeddings may have been previously created using a similar neural network.

[0022] Then, a vote is taken on the comparison result (e.g., comparison result 124) at 126. The vote produces an authentication result (e.g., authentication corresponds to touch input 106 from one of the authorized users or rejection of touch input 106 from one of the authorized users). If the authentication result is authentication (e.g., "yes" in 124), the computing device 102 can perform the action corresponding to touch input 106. However, if the computing device 102 is started in a locked state, additional actions can be performed (e.g., unlocking the computing device 102 and removing the soft lock screen). In the example shown, touch input 106 is an expand gesture. Therefore, the user interface can be scaled as shown at 128. If the authentication result is rejection (e.g., "no" in 124), the computing device 102 can be locked to prevent unauthorized use, as shown at 130. The lock can be a soft lock (e.g., the user interface is still displayed) or a hard lock (e.g., the user interface is replaced with a lock screen).

[0023] By performing such an action, the transition from locked to unlocked state can be more seamless, and users can be continuously authenticated through normal interaction with the device.

[0024] Example data stream

[0025] Figure 2 An example data flow 200 for a fingerprint module 202 used for fingerprint-based authentication using touch input is illustrated. Data flow 200 typically corresponds to at least a portion of process flow 100, and therefore may also be discussed below regarding... Figure 3 It is implemented in the computing device 102 (not shown) discussed.

[0026] A fingerprint module 202, implemented at least partially in the hardware of computing device 102, receives raw image data 112 corresponding to touch input 106 and provides authentication result 204. The raw image data 112 may include pixel data for the entire under-display fingerprint sensor or for a portion of the under-display fingerprint sensor corresponding to touch input 106. For example, computing device 102 may determine the general location of touch input 106 (e.g., via a capacitive sensor on the touchscreen) and retrieve raw image data corresponding to the location. Furthermore, to enhance the raw image data 112, computing device 102 may illuminate the location (e.g., via pixels on the touchscreen or backlight). Authentication result 204 corresponds to... Figure 1 The authentication results of the 126 votes (e.g., authentication corresponds to the touch input 106 of an authorized user or rejection of touch input 106 that does not correspond to the touch input 106 of an authorized user). Various actions can be performed based on the authentication results, as will be discussed further below.

[0027] The raw image data 112 is received by the clustering module 206 of the fingerprinting module 202 that generates a portion of the raw image data 118. To do this, the clustering module 206 can use a Gaussian mixture model 208 with preset clustering values ​​to determine corresponding portions of clusters 116 and the raw image data 118. The clustering module 206 can determine which pixels in the raw image data 112 have values ​​above a threshold (e.g., intensity, color, or brightness), determine the number of pixel clusters, and assign each pixel with a value above the threshold to one of the clusters 116. Therefore, cluster 116 is cluster 112 of the pixels in the raw image data—e.g., portions of the raw image data 118.

[0028] The clustering values ​​within Gaussian mixture model 208 (e.g., the number of Gaussian clusters) can be set to values ​​higher than the number of stored fingerprint embeddings (e.g., if four stored fingerprint embeddings exist, the number of clusters can be set to ten). The clustering values ​​can also be set to values ​​corresponding to the maximum expected number of touches in touch input 106. Since the likelihood of more than ten fingers or objects being associated with touch input 106 is small, ten values ​​can be chosen for the clustering values. Note that the clustering values ​​are part of the clustering technique (e.g., Gaussian mixture model 208) used to influence clustering, but do not directly indicate the number of clusters 116.

[0029] Although the discussion focuses on using Gaussian mixture model 208, clustering module 206 can utilize any clustering technique known to those skilled in the art to determine the corresponding portions of cluster 116 and the original image data 118. Furthermore, cluster values ​​within the selected clustering technique can be set based on experimental data (although those skilled in the art will recognize that setting values ​​greater than 10 may be counterproductive, as it may lead to the identification of touched portions as separate touches).

[0030] Once the Gaussian mixture model 208 (or another model) has determined the various parts of the original image data 118, the clustering module 206 can provide a portion of the original image data 118 to the embedding module 212 of the fingerprint module 202. If only one fingerprint is detected, the clustering module 206 will send a portion of the original image data 110 corresponding to that fingerprint.

[0031] Embedding module 212 utilizes neural network 214 with radial convolutional layer 216 to determine touch embedding 120 corresponding to a portion of the original image data 118 of touch 108. Radial convolutional layer 216 allows for rotational invariance of the portion of the original image data 118. In other words, radial convolutional layer 216 allows the associated touch to be in any orientation relative to the device. In this way, embedding module 212 does not require a specific fingerprint gesture to achieve usable embedding.

[0032] Neural network 214 may include additional elements known to those skilled in the art. For example, neural network 214 may include pooling (e.g., max pooling) and fully connected layers (e.g., normal layers) to determine touch embedding 120. As described above, touch embedding 120 may include hashes of respective portions of the original image data 118.

[0033] Neural network 214 has been trained using known inputs (e.g., weights have been computed). For example, multiple instances of fingerprint data of the same finger can be fed to neural network 214 at different times during normal use of the device (different parts of the fingerprint, different orientations of the fingerprint). Training can be repeated for multiple fingers and users to create a reliable network. By training neural network 214 in this way, neural network 214 is able to create similar embeddings regardless of the orientation or part of the associated touch (e.g., fingerprint). This is similar to facial recognition, where the system can identify a person even if part of their face may be obscured by glasses or a hat.

[0034] Once touch embeddings 120 have been created, comparison module 218 compares each of the touch embeddings 120 with a stored fingerprint embedding 220. The stored fingerprint embedding 220 is an embedding of an authorized user's fingerprint previously created using embedding module 212. For example, during the setup process, portions of one or more images corresponding to an authorized user's fingerprint can be fed into neural network 214 to create a stored fingerprint embedding. This process can then be repeated for other fingers and / or other authorized users.

[0035] The comparison may involve checking whether each touch embedding 120 matches or is sufficiently close to one of the stored fingerprint embeddings 220. For example, comparison module 218 may determine whether each touch embedding 120 is within a specific radius of the embedding space of any of the stored fingerprint embeddings 220. An example comparison using 1-sparse hits is shown below.

[0036] Example Comparison

[0037] This example comparison involves inputting one of the touch embeddings 120 via a query function (e.g., a 1-sparse hit function). The query function utilizes a forward-observation model and backward-inference computation to determine whether the touch embedding 120 matches any stored fingerprint embedding 220.

[0038] The forward model can be defined as:

[0039] y = Ax

[0040] Where y is a touch embedding 120 (e.g., a vector of a specific size), A is a set of stored fingerprint embeddings 220 (e.g., a matrix of a specific size consisting of multiple stored fingerprint embeddings 220), and x is a selection vector (e.g., representing a match) whose size is equal to the number of stored fingerprint embeddings 220.

[0041] Back-inference computation (1-sparse hit) can be defined as:

[0042]

[0043] The solution to the optimization problem computed by back-inference is essentially the similarity score between each of the touch embeddings 120 and the stored fingerprint embeddings 220. The optimization problem can be formulated as a sparse linear inverse system, which is solved using the following atomic matching principle:

[0044]

[0045]

[0046]

[0047] in, It consists of a touch-embedded 120 and a corresponding stored fingerprint-embedded 220 (A1, A2, A3...A...). n The index with the highest similarity score (dot product) between ) and f indicates if If the value is greater than the threshold T, then the stored fingerprint embedding 220 is a match.

[0048] If one of the similarity scores is higher than a threshold T, touch embedding 120 can be authenticated, and therefore touch 108 associated with touch embedding 120 can be authenticated. If none of the similarity scores is higher than the threshold T, touch embedding 120 can be rejected, and therefore touch 108 associated with touch embedding 120 can be rejected. In some embodiments, instead of or supplementing the similarity scores, the distance between touch embedding 120 and each stored fingerprint embedding 220 can be calculated.

[0049] After the comparison has been performed, the comparison result 124 is sent to the voting module 222 of the fingerprint module 202 to determine the authentication result 204 of the touch input 106. The comparison result 124 may include a set of binary integers corresponding to the corresponding authentication or rejection for each touch embedding 120. For example, if the touch 108 corresponds to the finger of an authorized user, the palm of an authorized user, and the finger of an unauthorized user, the set of binary integers would be [1, 0, 0].

[0050] Voting module 222 votes on comparison result 124. Voting may include determining whether the number or percentage of authenticated or rejected results within comparison result 124 exceeds a threshold. In the example above, comparison result 124 is [1,0,0], representing 33% authentication. If the threshold is 20%, then in this example, authentication result 204 would be authentication of touch input 106. However, if the threshold is 40%, then in this example, authentication result 204 would be rejection.

[0051] Although the fingerprint module is described as operating on each portion of the raw image data 118 to determine authentication result 204, in some embodiments, portions of the raw image data 118 may be operated on sequentially until a comparison result 124 for authentication is determined. For example, the fingerprint module 202 may select one of the portions of the raw image data 118, form a touch embedding 120 for the selected portion, and compare the touch embedding 120 with a stored fingerprint embedding 220. If one of the comparison results 124 is authentication, then the fingerprint module 202 may determine that the touch input 106 corresponds to an authorized user (e.g., providing authentication result 204 as authentication) without operating on the other portions of the raw image data 118. If no match is found, the fingerprint module 202 may sequentially select other portions of the raw image data 118, create associated touch embeddings 120, and compare those touch embeddings 120 with the stored fingerprint embeddings 220 until a comparison result 124 for authentication is determined. If the comparison result 124 does not contain any authentication, then the authentication result 204 will be a rejection; for example, the fingerprint module 202 will reject the touch input 106.

[0052] Example device

[0053] Figure 3 An example of a computing device 102 is illustrated in Figure 300, in which fingerprint-based authentication using touch input can be implemented. Although the computing device 102 is illustrated as a smartphone, it may include any electronic device (e.g., laptop, television, desktop computer, tablet, wearable device, etc.) without departing from the scope of this disclosure. As shown below, the computing device 102 includes at least one processor 302, at least one computer-readable storage medium 304, a touchscreen 306, a fingerprint imaging sensor 308, and a fingerprint module 202.

[0054] Processor 302 (e.g., application processor, microprocessor, digital signal processor (DSP), or controller) executes instructions 310 (e.g., code) stored in computer-readable storage medium 304 (e.g., non-transitory storage device, such as hard disk drive, SSD, flash memory, read-only memory (ROM), EPROM, or EEPROM) to cause computing device 102 to perform the techniques described herein. Instructions 310 may be part of an operating system and / or one or more applications of computing device 102.

[0055] Instruction 310 causes computing device 102 to take action on data 314 (e.g., application data, module data; sensor data, or I / O data) (e.g., create, receive, modify, delete, transmit, or display). Although shown as being within computer-readable storage medium 304, portions of data 314 may be located in random access memory (RAM) or cache of computing device 102 (not shown). Furthermore, instruction 310 and / or data 314 may be located remotely from computing device 102.

[0056] The fingerprint module 202 (or a portion thereof) may be included in or be a separate component of the computer-readable storage medium 304 (e.g., executed in dedicated hardware communicating with the processor 302 and the computer-readable storage medium 304). For example, instruction 310 may cause the processor 302 to either receive the raw image data 112 and output the authentication result 204, as per [reference to...]. Figure 1 and 2 As described.

[0057] The touchscreen 306 providing indication of touch input 106 and the fingerprint imaging sensor 308 providing raw image data 112 can be any of those known to those skilled in the art. For example, the touchscreen 306 can be a capacitive touchscreen, and the fingerprint imaging sensor 308 can be a full-display fingerprint sensor.

[0058] Example soft locking

[0059] Figure 4 This is an example illustration 400 of fingerprint-based authentication combined with a soft lock screen. Example illustration 400 shows that the computing device 102 begins to be in a locked state at 402 and displays the soft lock screen 404, and ends at 406 with an action corresponding to touch input 106.

[0060] At 402, computing device 102 is locked and a soft lock screen 404 is displayed. The soft lock screen 404 may be a blurred user interface, a user interface overlay (e.g., a pattern on the user interface), a color change, or other visual indicators that the computing device 102 is locked while still displaying the user interface.

[0061] In some implementations, computing device 102 may display a soft lock screen 404 in response to determining that user interaction is imminent. For example, if computing device 102 determines that it has been picked up or that the user is looking at touchscreen 306, the device can change from a hard lock screen to a soft lock screen with a suitable user interface (e.g., the last used app, the home screen, etc.). Step 402 typically occurs in... Figure 1 The described process is an example of what can be displayed before and when touch input 106 is received (e.g., in step 104).

[0062] Touch input 106 is received at 406, and touch input 106 is authenticated as corresponding to an authorized user (e.g., via...). Figure 1 and Figure 2 In the process (the process), the soft lock screen 404 is removed or changed into a user interface, and an action corresponding to touch input 106 is performed, as shown in 408. In the example shown, the user interface is scaled according to touch input 106. In this way, the user can interact with the computing device 102 without a separate unlocking step; for example, moving from 402 to 408 requires only a single user input.

[0063] Example Method

[0064] The following discussion describes a fingerprint-based authentication method using touch input. This method can be implemented using previously described examples, such as process flow 100, data flow 200, computing device 300, and... Figure 4 The illustration shown is 400. Figure 5 Aspects of method 500 illustrated herein are shown as operations 502 to 516 performed by one or more entities (e.g., computing device 102 and / or fingerprint module 202). The order in which the operations of the method are shown and / or described is not intended to be construed as limiting, and any number or combination of the described method operations may be combined in any order to implement the method or an alternative method.

[0065] Optionally, at 502, a soft lock screen is displayed. For example, touchscreen 306 may display a soft lock screen 404, which may be a blurred version of the underlying user interface screen or an overlay on top of the underlying user interface screen. Soft lock screen 404 indicates that computing device 102 is locked but can still be directly interacted with by an authorized user (e.g., no separate unlock input is required).

[0066] At 504, touch input is received. For example, touch input 106 may be received by touchscreen 306, and touch input 106 may include one or more touches 108. Touch input 106 may be a gesture and may be associated with the functionality of computing device 102. Furthermore, gestures may correspond to user interface elements displayed by computing device 102.

[0067] At 506, the original image data corresponding to the touch input is retrieved. For example, the fingerprint module 202 may retrieve or otherwise receive the original image data 112 corresponding to the touch input 106.

[0068] At 508, pixel clustering is performed on the raw image data. For example, clustering module 206 can use Gaussian mixture model 208 to determine the portions of the raw image data 118 corresponding to each touch 108.

[0069] At 510, a comparison result is established for the touch. The comparison result is established by forming a touch embedding representing the portion of the original image data corresponding to the corresponding touch for each touch, calculating the similarity between the touch embedding and each of one or more stored fingerprint embeddings corresponding to the fingerprints of one or more authorized users, and establishing a comparison result for the corresponding touch based on the calculation. For example, the embedding module 212 can use the radial convolutional layer 216 of the neural network 214 to form the touch embedding 120. The comparison module 218 can compare the touch embedding 120 with the stored fingerprint embedding 220 to establish a comparison result 124.

[0070] At 512, the authentication result is determined. For example, the voting module 222 can receive the comparison result 124 and determine the authentication result 204. The authentication result 204 can be an authentication corresponding to the touch input 106 of an authorized user or a rejection corresponding to the touch input 106 of an unauthorized user.

[0071] Optionally, at 514, an action is performed in response to an authentication result indicating that the touch input corresponds to an authorized user. For example, computing device 102 may determine that authentication result 204 indicates that touch input 106 corresponds to an authorized user, and perform an action corresponding to touch input 106 (e.g., zooming or selection of a user interface shown at 128).

[0072] Alternatively, at 516, the computing device is locked in response to an authentication result indicating that the touch input does not correspond to an authorized user. For example, computing device 102 may determine that authentication result 204 indicates that touch input 106 does not correspond to an authorized user and lock the device (e.g., a soft lock screen 404 or a lock screen depicted by 130).

[0073] Example

[0074] Example 1: A fingerprint-based authentication method performed by a computing device, the method comprising: receiving a touch input including one or more touches on a touchscreen of the computing device; retrieving raw image data corresponding to the one or more touches from a fingerprint imaging sensor of the computing device; performing a pixel clustering technique on the raw image data to determine one or more portions of the raw image data, each portion of the raw image data corresponding to one of the one or more touches; determining a portion of the raw image data corresponding to a corresponding touch based on the pixel clustering technique by establishing one or more comparison results of the touch input for each of the one or more touches; forming a touch embedding representing the portion of the raw image data corresponding to the corresponding touch; calculating a similarity between the touch embedding of the corresponding touch and each of one or more stored fingerprint embeddings corresponding to one or more authorized users' fingerprints; establishing the comparison results of the corresponding touch based on the calculation; and determining an authentication result of the touch input based on the one or more comparison results.

[0075] Example 2: The method of Example 1, wherein the authentication result for the touch input is further based on a vote on the one or more comparison results, the vote including: determining the number of the one or more comparison results corresponding to the authentication; and determining whether the number of the one or more comparison results corresponding to the authentication exceeds a threshold.

[0076] Example 3: The method of Example 1 or 2, wherein the fingerprint imaging sensor includes a fully display fingerprint imaging sensor.

[0077] Example 4: The method according to any of the preceding examples, wherein the touch input includes a single touch input.

[0078] Example 5: Any of the methods in the preceding examples, wherein the touch input is a plurality of touch inputs.

[0079] Example 6: A method of any of the foregoing examples: wherein the touch input is a gesture corresponding to a user interface; wherein the authentication result includes confirmation that the touch input corresponds to an authorized user; and further includes performing an action corresponding to the gesture.

[0080] Example 7: The method as described in any one of Examples 1 to 5: wherein the touch input is a gesture corresponding to a user interface; wherein the authentication result includes a determination that the touch input does not correspond to an authorized user; and further includes locking the computing device.

[0081] Example 8: Any of the methods in the foregoing examples further includes displaying a soft lock screen indicating that the computing device has been previously locked before receiving the raw image data.

[0082] Example 9: Any of the methods in the preceding examples, wherein the touch embedding and the one or more stored fingerprint embeddings include hashing.

[0083] Example 10: Any of the methods in the preceding examples, wherein the touch embedding and the one or more stored fingerprint embeddings are formed using radial convolutional layers of a neural network.

[0084] Example 11: Any of the methods in the preceding examples, wherein the calculated similarity corresponds to the radius distance between the touch embedding and the corresponding stored fingerprint embedding in the embedding space.

[0085] Example 12: Any of the methods in the foregoing examples, wherein the pixel clustering technique includes a hybrid model for determining one or more portions of the original image data.

[0086] Example 13: The method of Example 12, wherein the cluster value within the determined hybrid model affecting one or more portions of the original image data is set to a value less than or equal to 10.

[0087] Example 14: The method of Example 12 or 13, wherein the mixture model is a Gaussian mixture model that determines one or more portions of the original image data.

[0088] Example 15: A computing device comprising: a touchscreen; a fingerprint imaging sensor; at least one processor; and at least one computer-readable storage medium including instructions that, when executed by the at least one processor, cause the processor to perform any of the methods of the preceding examples.

[0089] Example 16: A computer-readable storage medium containing instructions that, when executed by one or more processors, cause the one or more processors to perform a method of any one of Examples 1-14. The computer-readable storage medium in this example may be a transient or non-transitory computer-readable storage medium.

[0090] While embodiments of fingerprint-based authentication using touch input have been described in language specific to certain features and / or methods, the subject matter of the appended claims is not necessarily limited to the specific features or methods described. Rather, the specific features and methods are disclosed as exemplary embodiments of fingerprint-based authentication using touch input. Furthermore, although various examples have been described above, each with certain features, it should be understood that a particular feature of one example is not necessarily specific to that example. Rather, any feature that complements or substitutes for any of those examples, any feature depicted above and / or in the figures, may be combined with any example.

Claims

1. A method for fingerprint-based authentication performed by a computing device, the method comprising: Receive touch input including one or more touches on the touchscreen of the computing device; Retrieve raw image data corresponding to the one or more touches from the fingerprint imaging sensor of the computing device; Perform pixel clustering on the original image data to determine one or more portions of the original image data. The pixel clustering technique includes clustering the pixels of the original image data into one or more clusters. The corresponding clusters in the one or more clusters correspond to the corresponding portions of the original image data of the one or more portions of the original image data. The corresponding portions of the original image data of the one or more portions of the original image data correspond to the corresponding touches in the one or more touches. One or more comparison results of the touch input are established by performing the following steps for each of the one or more touches: Using the pixel clustering technique that clusters the pixels of the original image data into one or more clusters, a corresponding part of the original image data corresponding to a corresponding touch is determined, wherein the corresponding cluster in the one or more clusters corresponds to the corresponding part of the original image data; Forming a touch embedding representing the portion of the original image data corresponding to the corresponding touch; Calculate the similarity between the touch embedding of the corresponding touch and each of one or more stored fingerprint embeddings corresponding to the corresponding fingerprints of one or more authorized users; as well as Based on the calculation, one or more comparison results are established for the corresponding touch; as well as In response to establishing the one or more comparison results: The touch input is authenticated as at least one corresponding fingerprint of a corresponding authorized user among the one or more authorized users; or Reject the touch input that does not correspond to the fingerprint of the one or more authorized users.

2. The method according to claim 1, wherein, The authentication or rejection is further based on a vote on one or more comparison results, the vote including: Determine the number of the one or more comparison results corresponding to the authentication; and Determine whether the number of the one or more comparison results corresponding to the authentication exceeds a threshold.

3. The method according to claim 1, wherein, The fingerprint imaging sensor includes a full-display fingerprint imaging sensor.

4. The method according to claim 1, wherein, The touch input includes single-touch input.

5. The method according to claim 1, wherein, The touch input is a multi-touch input.

6. The method according to claim 1: in, The touch input is a gesture corresponding to the user interface; Wherein, the touch input corresponds to at least one corresponding fingerprint of the corresponding authorized user; and The method further includes performing an action corresponding to the gesture.

7. The method according to claim 1: in, The touch input is a gesture corresponding to the user interface; Wherein, the touch input does not correspond to the corresponding fingerprint of the one or more authorized users; and The method further includes locking the computing device.

8. The method of claim 1, further comprising displaying a soft lock screen indicating that the computing device has been previously locked before receiving the raw image data.

9. The method according to claim 1, wherein, The touch embedding and the one or more stored fingerprint embeddings include hashes.

10. The method according to claim 1, wherein, The touch embedding and the one or more stored fingerprint embeddings are formed using radial convolutional layers of a neural network.

11. The method according to claim 1, wherein, The calculated similarity corresponds to the radius distance between the touch embedding and the corresponding stored fingerprint embedding in the embedding space.

12. The method according to any one of claims 1-11, wherein, The pixel clustering technique includes a hybrid model that determines one or more portions of the original image data.

13. The method according to claim 12, wherein, The clustering values ​​within the determined hybrid model that affect one or more portions of the original image data are set to values ​​less than or equal to 10.

14. The method according to claim 12, wherein, The mixture model is a Gaussian mixture model that determines one or more portions of the original image data.

15. A computing device, comprising: touchscreen; Fingerprint imaging sensor; At least one processor; as well as At least one computer-readable storage medium including instructions that, when executed by the at least one processor, cause the at least one processor to perform the method of any one of claims 1-14.

Images