Facial feature recognition

Abstract

Procedure that includes: Storing a set of templates (60) of removable facial features and a set of templates (62) of non-removable facial features; Weights of each of the plurality of templates (60) of removable facial features and the plurality of templates (62) of non-removable facial features such that the facial feature templates that have been incorrectly classified by a given classifier are more likely to be selected for further training than facial feature templates that have been correctly classified; Capturing an image containing at least one user's face (30) by a computer device (10); Calculating a face template of the user's face (30) in the image using the computer device (10); Determine whether the face contains a removable facial feature that reduces a degree of dissimilarity between two faces, by the computer device (10) and at least partially on the basis of a weighted sum of similarity assessments between the face template and each of the plurality of templates (60) of removable facial features; Determine whether the face contains a non-removable facial feature that reduces a degree of dissimilarity between two faces, by the computer device (10) and at least partially on the basis of a weighted sum of similarity assessments between the face template and each of the plurality of templates (62) of non-removable facial features; wherein the procedure, if the face contains the removable facial feature, comprises: issuing a message to the user (30) that he should remove the removable facial feature, by means of the computer device (10); and wherein the procedure, if the face contains the non-removable facial feature, comprises: setting a first similarity assessment threshold to a second similarity assessment threshold by the computer device (10).

Description

TECHNICAL AREA

[0001] This disclosure relates to a facial recognition technology and, in particular, to the capture of facial features from images taken for use in facial recognition. BACKGROUND

[0002] A user can activate a computer device by "unlocking" the device or otherwise gain access to its functionalities. In some cases, a computer device may be configured to allow unlocking based on authentication information provided by the user. Authentication information can take various forms, including alphanumeric access codes, gestures, and biometric information. Examples of biometric information include fingerprints, retinal scans, speech, and facial images. A computer device can authenticate facial image input using facial recognition technology. The following publications describe various aspects of biometric information processing: John Gill, "The Markets for the Adaptation of Self-Service Terminals to be Accessible by People with Disabilities," European Commission, 2009 (http: / / ec.europa.eu / information_society / activities / einclusion / docs / worshop_atm / atm _markets_report.doc); US 7817826 B2; DE 102009023306 A1; Kumar, N., Berg, A.C., Belhumeur, P.N., Nayar, S.K. „Describable Visual Attributes for Face Verification and Image Search“ Proceedings of the IEEE, 1995, Volume 83 , Issue 5, Pages 705 - 741. ZUSAMMENFASSUNG

[0003] One objective of the present disclosure is to improve facial recognition technology. This objective is achieved by the method according to claim 1. Further embodiments are defined in the dependent claims. In one example, a method comprises capturing an image containing at least one user's face by a computer device and calculating a facial template of the user's face in the image by the computer device. The method further comprises analyzing the facial template to determine whether the face contains a removable facial feature that reduces the degree of distinctiveness between two faces, and / or a non-removable facial feature that reduces the degree of distinctiveness between two faces.If the face contains the removable facial feature, the method further includes the computer device issuing a message to the user to remove the removable facial feature. If the face contains the non-removable facial feature, the method further includes the computer device setting a first similarity assessment threshold to a second similarity assessment threshold.

[0004] In another example, a computer-readable storage medium contains instructions to cause at least one processor of a computer device to perform operations. These operations include capturing an image containing at least one user's face and calculating a face template of the user's face in the image. Furthermore, the operations include analyzing the face template to determine whether the face contains a removable facial feature, which reduces the degree of distinctiveness between two faces, and / or a non-removable facial feature, which also reduces the degree of distinctiveness between two faces. If the face contains the removable facial feature, the operations further include issuing a message to the user instructing them to remove the removable facial feature.If the face contains the non-removable facial feature, the operations further include adjusting one similarity assessment threshold to a second similarity assessment threshold.

[0005] In another example, a device includes at least one processor, at least one camera which can be operated by the at least one processor to take a picture, and at least one output device.The at least one processor is configured to analyze the first face template to determine whether the face contains a removable facial feature that reduces a degree of dissimilarity between two faces, and / or a non-removable facial feature that reduces a degree of dissimilarity between two faces, wherein the at least one output device is configured to issue a message to the user to remove the removable facial feature if the face contains the removable facial feature, and wherein the method includes setting a first similarity assessment threshold to a second similarity assessment threshold if the face does not contain the non-removable facial feature.

[0006] In another example, a method includes a computer device capturing an image containing at least one user's face and calculating a face template of the user's face in the image. The method further includes the computer device analyzing the face template to determine whether the face contains a removable facial feature that reduces the degree of distinctiveness between two faces. If the face contains the removable facial feature, the method further includes the computer device issuing a message to the user instructing them to remove the removable facial feature.

[0007] The details of one or more examples are set forth in the accompanying drawings and in the following description. Further features and advantages will become apparent from the description, the drawings, and the claims. BRIEF DESCRIPTION OF THE DRAWINGS Fig. Figure 1 is a conceptual diagram illustrating an exemplary computer device that captures facial features which can reduce the degree of dissimilarity between two faces during facial recognition, in accordance with one or more aspects of the present disclosure. Fig. Figure 2 is a block diagram showing details of an exemplary computer device for capturing facial features which can reduce the degree of dissimilarity between two faces during facial recognition, in accordance with one or more aspects of the present disclosure. Fig. Figure 3 is a conceptual diagram illustrating an exemplary computer device that captures removable facial features which can reduce the degree of dissimilarity between two faces during facial recognition, in accordance with one or more aspects of the present disclosure. Fig. Figure 4 is a conceptual diagram illustrating an exemplary computer device that captures non-removable facial features which can reduce the degree of dissimilarity between two faces during facial recognition, in accordance with one or more aspects of the present disclosure. Fig. 5 is a flowchart that represents an exemplary process that can be carried out by a computer device to analyze a facial image and to capture facial features that can reduce a degree of dissimilarity between two faces during facial recognition, in accordance with one or more aspects of the present disclosure. Fig. 6 is a flowchart that represents an exemplary process that can be carried out by a computer device when a removable facial feature has been captured in the image for facial recognition, which can reduce a degree of dissimilarity between two faces, in accordance with one or more aspects of the present disclosure. Fig. 7 is a flowchart that represents an exemplary process that can be carried out by a computer device when a non-removable facial feature, capable of reducing a degree of dissimilarity between two faces, has been captured in the image for facial recognition, in accordance with one or more aspects of the present disclosure. DETAILED DESCRIPTION

[0008] A computer device can use facial recognition programs in various scenarios. For example, a computer device can use facial recognition programs to authenticate a user attempting to access one or more functionalities of the computer device, or functionalities otherwise controlled by the computer device. In some common scenarios, a computer device can store templates (or "registered templates") calculated from images of the faces of one or more authorized users (or "registration images"). When a user attempts to access (or "unlock") functionalities of the computer device, the computer device can capture an image of the user's face for authentication purposes. The computer device can calculate a template of the captured facial image and apply facial recognition techniques (e.g.,The computer uses a facial recognition program to compare the captured facial image template with registered templates assigned to authorized users. If the facial recognition program determines an acceptable degree of match between the template and at least one registered template, the computer device can authenticate the user and grant the access request.

[0009] In general, a facial recognition program can authenticate a user and grant them access to the computer device's functionalities by comparing facial features of the captured image with those of the registration images assigned to authorized users. If the facial features of the captured image are sufficiently similar to those of one of the registration images, the user is authorized to access the computer device's functionalities. The similarity between facial features can be quantified using a similarity score, with higher scores indicating a greater degree of similarity. If a similarity score when comparing the facial features of two images is higher than a similarity score threshold, the facial features of the two images can be considered a match.

[0010] However, facial recognition authentication may not function as intended in some cases. For example, certain facial features, such as sunglasses and facial hair, can cause facial recognition authentication to fail. As discussed here, one example of such a failure is when a user is denied a match even though they are an authorized user. Another example is when a user is granted a match when they are not an authorized user.

[0011] Authorized users can become frustrated when they are denied access to the computer device on which they are indeed an authorized user. Furthermore, unauthorized users can exploit vulnerabilities in facial recognition programs that can lead to incorrect authentication. For example, an unauthorized user might attempt to unlock a computer device by trying to appear less distinctive. If they appear less distinctive, this could cause a facial recognition system to mistake a captured image of an unauthorized user for one of the registered images associated with authorized users.

[0012] In some examples, facial features that can reduce the degree of dissimilarity between users can be any facial feature that increases a similarity score between two otherwise different users. In other words, a facial feature that can reduce the degree of dissimilarity between users is a facial feature that causes a similarity score for two images of different people to be increased because of that facial feature. In some examples, facial features that can reduce the degree of dissimilarity between users can be classified into two categories: removable facial features and non-removable facial features. For example, removable facial features that can reduce the degree of dissimilarity between users might include objects such as sunglasses, headgear (e.g., hats), or other head coverings.Non-removable facial features may include, but are not limited to, hats and other objects that can obscure part of a user's face. Furthermore, non-removable facial features that can reduce the degree of diversity between users may include, but are not limited to, facial hair. Examples of facial hair include goatees, sideburns, mustaches, beards, and costume facial hair (e.g., fake facial hair). Non-removable facial features may include features that are not easily removable because they are, for example, physically attached to a user, but which can eventually be removed with additional effort (e.g., by actions such as shaving the facial hair).

[0013] In some examples, a user's eyes can be one or more distinguishing features between users. A user wearing sunglasses covers their eyes and can thus reduce the degree of distinctiveness between users. Similarly, facial hair can conceal or obscure various distinguishing features of a user and thus reduce the degree of distinctiveness between users. That is, two otherwise different users may tend to look more similar if they wear sunglasses and / or have facial hair.

[0014] In some examples, an unauthorized user possessing either removable and / or non-removable facial features may erroneously gain access to a computer device. For instance, a facial recognition system might calculate a similarity score that exceeds a similarity threshold for a captured image of the unauthorized user, allowing them to access the computer device. Thus, unauthorized users may attempt to appear less distinctive in order to gain access to the computer device's functionalities and circumvent authentication restrictions otherwise implemented by the device.

[0015] In general, the present disclosure relates to techniques for preventing facial recognition failures by analyzing a facial template calculated from a captured image for removable and non-removable facial features, which can reduce the degree of dissimilarity between two otherwise different images. In some examples, the techniques can be executed by programs (e.g., facial recognition programs) running on the computer device that can detect the removable and non-removable facial features, thus reducing the degree of dissimilarity between two faces. Detecting the removable and non-removable facial features can reduce instances of failed authentication and / or reduce instances in which an authorized user is denied access to the computer device.

[0016] In one example, facial recognition programs can pause the facial authentication process and generate a message prompting the user to remove the removable facial feature if a removable feature is detected in the facial template. In some examples, the message can instruct the user to take a subsequent image (e.g., an image without the removable facial feature) for facial authentication. In another example, facial recognition programs can increase a similarity assessment threshold to a predefined level if a non-removable facial feature is detected. In some examples, facial recognition programs can generate a message indicating that security is being increased and prompting the user to choose whether to continue with facial recognition.In some examples, adjusting (e.g., increasing) the similarity assessment threshold can compensate for the reduction in the degree of dissimilarity between a captured image and a registered image due to the non-removable facial feature.

[0017] In another example, removable and non-removable facial features, which can reduce the degree of dissimilarity between two faces, can be captured during a registration process. For example, the facial recognition program can provide a notification indicating that the potential registration template contains the removable and / or non-removable facial feature if a removable and / or non-removable facial feature is captured in a potential registration template calculated from a registration image. Furthermore, the facial recognition program can adjust the similarity assessment threshold for the registration template containing the removable and / or non-removable facial feature.In other examples, facial recognition programs can prevent the registration template from being saved as a registration template of an authorized user and prompt the user to take another picture for registration purposes.

[0018] The facial recognition programs described in this disclosure can mitigate the risks of an unauthorized user causing an erroneous authorization by attempting to appear less distinctive, and can mitigate the risks of denied access to authorized users. For example, the facial recognition programs can deny a user authorization if the object of facial recognition includes a removable facial feature (e.g., sunglasses) that can reduce the degree of distinctiveness. In another example, the facial recognition programs can adjust the security level (e.g., increase a similarity assessment threshold) if the object of the facial image includes non-removable facial features (e.g., a beard) that can reduce the degree of distinctiveness.In this way, techniques of the present disclosure can capture facial features, which can reduce the degree of distinctiveness between two faces. Furthermore, facial recognition results can be improved by preventing potential registration templates containing removable and / or non-removable facial features from being stored as registration templates.

[0019] Fig. Figure 1 is a conceptual diagram illustrating an exemplary computer device that captures facial features which can reduce the degree of dissimilarity between two faces during facial recognition, in accordance with one or more aspects of the present disclosure. Fig. Figure 1 illustrates a computer device 10 that can capture a facial image associated with a user (e.g., user 30) and determine whether the captured facial feature contains a feature that can reduce the degree of distinctiveness between two faces during facial recognition. In the example from Fig. 1. The user 30 may hold a computer device 10 (e.g., a mobile computer device) that can perform facial recognition based on one or more images of the user 30 captured by the computer device 10. The computer device 10 may include, be, or be part of one or more of a variety of device types, such as, but not limited to, a mobile phone (including so-called "smartphones"), tablet computer, netbook, laptop, desktop computer, personal digital assistant ("PDA"), set-top box, television, and / or watch.

[0020] The computer device 10 includes one or more input devices 24, at least one of which is a camera 26. The camera 26 can be part of a forward-facing camera of the computer device 10 or coupled to it. In other examples, the camera 26 can be part of a rear-facing camera of the computer device 10 or coupled to it. The forward-facing and / or the rear-facing camera can be capable of capturing still images, video, or both.

[0021] In the example from Fig. The computer device 10 further includes output devices 28. At least one of the output devices 28 can display a graphical user interface (GUI) 32. The GUI 32 can be displayed by a variety of display devices, including input / output devices such as a touchscreen or presence-sensitive displays. In various examples, the computer device 10 can cause one or more output devices 28 to update the GUI 32 so that it contains various user interface controls, text, images, or other graphic content. Displaying or updating the GUI 32 can generally refer to the process of causing one or more output devices 28 to change the content of the GUI 32 that can be displayed to the user.

[0022] In the example from Fig. 1. User 30 may not currently be authenticated to computer device 10. However, user 30 may wish to be authenticated to computer device 10. User 30 may authenticate to computer device 10 using facial recognition techniques. The GUI 32 may display graphical information relating to the authentication of a user to computer device 10 using facial recognition in accordance with the techniques of this disclosure. As in the example from Fig. As shown in Figure 1, the GUI 32 can contain one or more GUI elements such as a backup indicator 34, an unbackup prompt 36 and a recording icon 38.

[0023] The computer device 10 can be configured to operate in a "safe" mode, represented by the presence of the safety indicator 34. In some examples, a user can actively configure the computer device 10 to operate in the safe mode. For example, the user can press a button (such as a safety button) or an icon displayed by the computer device 10 for a predetermined duration to request that the computer device 10 operate in the safe mode before entering it. In these and other examples, a user can tap, swipe, or otherwise interact with one or more elements of the GUI 32 using an output device 28 (such as a presence-sensitive screen) of the computer device 10.Furthermore, the computer device 10 can be passively configured to operate in secure mode. For example, a predefined period of "inactivity" can configure the computer device 10 to operate in secure mode. Inactivity can occur due to the absence of user interaction (e.g., button presses, contact with input devices 24 and / or output devices 28, etc.). The predefined period that configures the computer device 10 to operate in secure mode can be a standard duration specified by a manufacturer of the computer device 10 or can be programmed by an authorized user, such as user 30.

[0024] In some examples, the computer device 10 can use facial recognition technology to halt operation in secure mode. In other words, the user 30 can "unsecure" the computer device 10 through authentication procedures that use facial recognition techniques to determine whether the user 30 is an authorized user of the computer device 10. More specifically, the user 30 can register with a facial recognition application or an embedded process of the computer device 10 by storing one or more user-registered templates 22 (e.g., templates calculated from registration images) that represent the user 30's face. The user 30 can cause the camera 26 of the computer device 10 to capture one or more user registration images.The computer device 10 can calculate user-registered templates 22 from the user registration images and can store the one or more user-registered templates 22 in one or more storage devices of the computer device 10 (e.g., in the facial recognition database 18) and / or at a remote location commonly known as a "cloud storage" for later use during facial recognition.

[0025] To unlock the computer device 10 using facial recognition technology, the user 30 can provide an authentication image that represents at least part of their face. In some examples, the user 30 can actively cause the camera 26 of the computer device 10 to capture the authentication image. For example, the user 30 can face a camera lens associated with the camera 26 and press a button to cause the camera 26 to capture the authentication image. In another example, the user 30 can tap, swipe, or otherwise interact with an area associated with the capture icon 38 contained in the GUI 32. In still other examples, the computer device 10 can automatically capture the authentication image in response to the fact that the user 30 is facing a camera lens associated with the camera 26. As in the example from Fig. As shown in Figure 1, the computer device 10 can display a GUI 32 containing a decryption prompt 36. In this example, the decryption prompt 36 indicates that the user 30 can simply face the camera 26, which may contain a camera lens or be coupled to it in some other way, to cause the computer device 10 to capture the authentication image.

[0026] As in Fig. As shown in Figure 1, the computer device 10 can also include a face recognition module 12 and a face recognition database 18. In one example, the face recognition module 12 can determine whether an authentication image for face recognition contains a removable facial feature, which can reduce the degree of distinctiveness between two users, and / or a non-removable facial feature, which reduces the degree of distinctiveness between two users.

[0027] In some examples, the facial recognition database 18 can have a logical and / or physical location (e.g., a logical location that refers to a specific physical location) of one or more storage devices or a memory (e.g., the one in Fig. The facial recognition database 18 may comprise the storage devices 58 shown in Figure 2 and the memory 52 shown therein. In some examples, the facial recognition database 18 may comprise a directory or file of a file system, a database, or a sector or block of a hard disk drive, a solid-state drive, or flash memory. In some examples, the facial recognition database 18 may also reside, at least partially, in the memory 52 of the computer device 10 (e.g., if the images of the facial recognition database 18 are cached on one or more storage devices 58 for later write-through).

[0028] In one example, the facial recognition module 12 can begin a facial recognition process after the computer device 10 has captured the authentication image and calculated an authentication face template from it. In some examples, the facial recognition process can take place in two phases: a facial feature acquisition phase and an authentication phase. During the facial feature acquisition phase, the facial feature module 14 can determine whether the authentication image contains any of the removable and / or non-removable facial features that can reduce the degree of distinctiveness between two users.

[0029] As discussed here, the facial feature module 14 can compare the authentication face template calculated from the user's authentication image 30 with several facial feature templates 20 stored in the facial recognition database 18 to determine whether the authentication face template contains a removable and / or a non-removable facial feature that can reduce the degree of distinctiveness between two users. The facial recognition module 12 can, for example, contain images from stock photo agencies (e.g., facial feature templates 20). That is, the facial feature templates 20 can be reference templates for a particular feature (e.g., templates 60 for removable facial features and templates 62 for non-removable facial features, as in Fig. (as shown in Figure 2). As discussed here, the facial feature templates can contain 20 templates calculated from images of individuals with and without the removable and non-removable facial features that can reduce the degree of dissimilarity between two users. Each template of the facial feature templates 20 can contain a normalized similarity score between the template and a facial template for which the presence of the removable and / or non-removable feature is known.

[0030] In some examples, the facial feature module 14 can compare the authentication face template with the facial feature templates 20 and calculate a similarity score. The facial feature module 14 can, for example, use facial recognition methods (such as facial recognition algorithms) to calculate the similarity scores. Subsequently, for each comparison between the authentication face template and the facial feature templates, the facial feature module 14 can determine a weighted sum of the calculated similarity scores. Based on this weighted sum of calculated similarity scores, the facial feature module 14 can determine whether the authentication image contains the removable and / or non-removable facial features that can reduce the degree of dissimilarity between two users.

[0031] Although the examples of disclosure are mainly described in relation to the capture of removable facial features such as sunglasses and non-removable facial features such as facial hair, the examples are not limited to these. Rather, some examples can be used to capture any facial feature that can reduce the degree of distinctiveness between two users.

[0032] During the authentication phase, the computer device 10 can compare the authentication template with the templates 22 registered by the user and determine whether the images are sufficiently similar for facial recognition purposes. If the facial authentication module 16 of the computer device 10 determines that the authentication facial template is sufficiently similar to one or more of the templates 22 registered by the user, the computer device 10 can grant the user 30 access to functionalities and content of the computer device 10. If the computer device 10 determines that the features of the authentication facial template do not match those of the template 22 registered by the user in the facial recognition database 18, the computer device 10 can deny the user 30 access to the functionalities and content of the computer device 10.

[0033] In some examples, the computer device 10 can use one or more facial recognition programs to compare metrics associated with the authentication face template with metrics associated with the templates 22 registered by the user. Some examples of metrics may include distances between facial features (pupil to pupil, mouth width, etc.), the outlines of various facial features, pixelation according to skin tone or texture, hair and / or eye color, and many others.The facial recognition programs executed in the computer device 10 can perform the comparison using one or more well-known recognition algorithms, such as geometric and / or photometric approaches, three-dimensional modeling techniques (3D modeling techniques) and recognition techniques, principal component analysis using eigensurfaces, linear decision analysis, elastic bundle graph equality checking, pattern equality checking, and dynamic link equality checking, to name just a few. Based on comparison-based values, such as pre-programmed acceptable margins of error, the facial recognition programs executed in the computer device 10 can determine whether the authentication face template and at least one user-registered face recognition template 22 are sufficiently similar to each other.In cases where the facial recognition programs recognize a match, user 30 can successfully unlock the computer device 10. Conversely, if the facial recognition programs reject a match, user 30 may not be able to unlock the computer device 10, and the computer device 10 may continue to operate in secure mode.

[0034] As discussed here, the similarity between the authentication face template and the templates 22 registered by the user is quantified using a similarity score, with higher similarity scores indicating a higher degree of similarity. If a similarity score between the authentication face template and one of the templates 22 registered by the user is higher than a similarity score threshold, the two images are considered to match, and the user 30 may be able to unlock the computer device 10. Conversely, if the similarity score between the authentication face image and the template 22 registered by the user is lower than the similarity score threshold, the two images are not considered to match, and the computer device 10 may prevent the user 30 from unlocking the computer device 10.

[0035] The computer device 10 can implement techniques of the present disclosure to determine whether an authentication image contains a removable facial feature that can reduce the degree of distinctiveness between two faces, and / or a non-removable facial feature that can reduce the degree of distinctiveness between two faces. If the computer device 10 determines that an authentication image contains a removable facial feature, the computer device 10 can also implement techniques of the present disclosure to issue a message 30 to the user to remove the removable facial feature and to take a second image for authentication.Furthermore, the computer device 10 can implement techniques of the present disclosure for increasing a similarity assessment threshold to a set similarity assessment threshold if the computer device 10 determines that an authentication image contains a non-removable facial feature.

[0036] Fig. Figure 2 is a block diagram detailing an exemplary computer device for capturing facial features that can reduce the degree of distinctiveness between two faces during facial recognition, in accordance with one or more aspects of the present disclosure. The computer device 10 may be a non-limiting example of the computer device 10 from Fig. 1. In other cases, many other exemplary embodiments of the computer device 10 can be used.

[0037] As in the example from Fig. As shown in Figure 2, the computer device 10 includes one or more processors 50, a memory 52, one or more storage devices 58, one or more input devices 24, one or more output devices 28, a network interface 54 and a camera 26. Components of the computer device 10 can be connected to each other (physically, communication-wise and / or functionally) for communication between components.

[0038] In some examples, one or more processors 50 are configured to implement functionality and / or process instructions for execution within the computer device 10. For example, the processors 50 can process instructions stored in memory 52 and / or instructions stored on storage devices 58. These instructions can include components of the operating system 64, the face recognition module 12, which contains the registration module 68, the face feature module 14, and the face authentication module 16, and one or more applications 66. Furthermore, the computer device 10 can contain one or more Fig. 2 additional components not shown, such as a power supply (e.g. a battery), a global positioning system (GPS) receiver and a high-frequency identification (RFID) reader.

[0039] In one example, the memory 52 is configured to store information within the computer device 10 during operation. In some examples, the memory 52 is described as a computer-readable storage medium. In some examples, the memory 52 is short-term storage, meaning that a primary purpose of the memory 52 cannot be long-term storage. In some examples, the memory 52 is described as volatile memory, meaning that the memory 52 does not retain any stored contents when it is not receiving power. Examples of volatile memory include read / write memory (RAM), dynamic read / write memory (DRAM), static read / write memory (SRAM), and other forms of volatile memory known in the field. In some examples, the memory 52 is used to store program instructions for execution by processors 50. In one example, the memory 52 is accessed by software (e.g.,used by the operating system 64) or by applications (e.g., by one or more applications 66) that are run in the computer device 10 to temporarily store information during program execution.

[0040] In some examples, one or more storage devices 58 also contain one or more computer-readable storage media. In some examples, the storage devices 58 may be configured to store larger amounts of information than the memory 52. ​​Furthermore, the storage devices 58 may be configured for long-term storage of information. In some examples, the storage devices 58 contain non-volatile storage elements. Examples of such non-volatile storage elements include magnetic disks, optical disks, solid-state drives, floppy disks, flash memory, forms of electrically programmable memory (EPROM) or electrically erasable and programmable memory, and other forms of non-volatile memory known in the field.

[0041] As in Fig. As shown in Figure 2, the computer device 10 may also include one or more input devices 24. One or more input devices 24 may be configured to receive input from a user via tactile, audio, video, or biometric channels. Examples of input devices 24 may include a keyboard, a mouse, a touchscreen, a presence-sensitive display, a microphone, one or more still and / or video cameras, a fingerprint reader, a retinal scanning device, or any other device capable of capturing input from a user or other source and transmitting the input to the computer device 10 or components thereof. Although the camera 26 in Fig. 2 shown separately, it may in some cases be part of the input devices 24.

[0042] In some examples, the output devices 28 of the computer device 10 can be configured to provide output to a user via visual, auditory, or tactile channels. The output devices 28 can include a video graphics adapter card, a liquid crystal display monitor (LCD monitor), a light-emitting diode monitor (LED monitor), a cathode ray tube monitor (CRT monitor), a sound card, a loudspeaker, or any other device capable of producing output that can be understood by a user. Furthermore, the output devices 28 can include a touchscreen, a presence-sensitive display, or other input / output-capable displays known in the field.

[0043] In some examples, the computer device 10 also includes a network interface 54. In one example, the computer device 10 uses the network interface 54 to communicate with external devices over one or more networks, such as one or more wireless networks. The network interface 54 can be a network interface card, such as an Ethernet card, an optical transceiver, a radio frequency transceiver, or any other type of device capable of sending and receiving information. Other examples of such network interfaces include Bluetooth, 3G, 4G, and WLAN radio in computer devices, as well as USB. In some examples, the computer device 10 uses the network interface 54 to communicate wirelessly with external devices over a network.

[0044] The operating system 64 can control one or more functionalities of the computer device 10 and / or components thereof. For example, the operating system 64 can interact with applications 66 and enable one or more interactions between applications 66 and one or more processors 50, a memory 52, storage devices 58, input devices 24, and output devices 28. As in Fig. As shown in Figure 2, the operating system 64 can interact with applications 66 and with a face recognition module 12 and its components, or be coupled with them in other ways. In some examples, a registration module 68 and / or a face feature module 14 and / or a face authentication module 16 can be included in the operating system 64. In these and other examples, the registration module 68 and / or the face feature module 14 and / or the face authentication module 16 can be part of applications 66. In other examples, the registration module 68 and / or the face feature module 14 and / or the face authentication module 16 can be implemented externally to the computer device 10, such as at a location on the network.In some such cases, the computer device 10 can use the network interface 54 to access the facial recognition module 12 and its components via methods commonly known as “cloud computing” and to implement functionalities provided by it.

[0045] The facial recognition module 12 can implement one or more techniques described in this disclosure. For example, the facial recognition module 12 can be configured to determine whether an authentication image captured by the camera 26 contains a removable facial feature that can reduce the degree of distinctiveness between two faces, and / or a non-removable facial feature that can reduce the degree of distinctiveness between two faces. Furthermore, the facial recognition module 12 can be configured to allow or deny access to the computer device 12 by identifying an authorized user of the computer device 10.

[0046] A user can interact with the computer device 10 to cause the computer device 10 to enter a registration mode, allowing an authorized user to register facial features with the computer device 10. During the registration process, the input device 24, such as the camera 26, can capture one or more registration images of the authorized user, including one or more registration images of the authorized user's face. The registration module 68 can calculate a registered face template from the registration images and store the templates as user-registered templates 22 in the facial recognition database 18. The registered face templates 22 can be a statistical representation of a face's appearance. The registration module 68 can, for example,Various facial features, such as the nose, eyes, mouth, nostrils, chin, forehead, eyebrows, cheekbones, and the like, are extracted, including but not limited to properties such as the location, size, and relationship between these features. Facial features extracted by the registration module 68 may also include facial properties such as lines, curves, edges, points, areas, and the like. The facial features extracted from the one or more images of the authorized user's face can be stored in the computer device 10 as registered face templates 22, which can be used during a facial recognition authentication process. As discussed here, more than one user may be authorized to access a computer device. Thus, the templates 22 registered by the user may contain registered face templates 22 of one or more authorized users.

[0047] In some examples, storing the registered face templates 22 may include storing one or more values ​​representing the facial features of each authorized user. The face authentication module 16 can retrieve the facial features stored in the user-registered templates 22, so that the face authentication module 16 can compare the facial features with facial features extracted from one or more templates calculated from images captured by the camera 26 to determine whether a user attempting to access the computer device 10 is an authorized user of the computer device 10.

[0048] Furthermore, the facial recognition database can contain 18 facial feature templates. As discussed here, the facial feature templates can contain 60 removable facial feature templates and 62 non-removable facial feature templates. Removable facial feature templates can contain one or more positive templates of a removable facial feature and one or more negative templates of a removable facial feature. The one or more positive templates of a removable facial feature are templates calculated from images of individuals possessing the removable facial feature, which can reduce the degree of dissimilarity between two faces.The one or more negative templates of a removable facial feature are templates calculated from images of people who do not possess the removable facial feature, which can reduce the degree of dissimilarity between two faces. In an example, the removable facial feature is sunglasses. Thus, the one or more positive templates of a removable facial feature are templates calculated from images of people wearing sunglasses, and the one or more negative templates of a removable facial feature are templates calculated from images of people not wearing sunglasses.

[0049] The templates for 62 non-removable facial features can contain one or more positive templates of a non-removable facial feature and one or more negative templates of a non-removable facial feature. The one or more positive templates of a non-removable facial feature are templates calculated from images of individuals who possess the non-removable facial feature, which can reduce the degree of dissimilarity between two faces. The one or more negative templates of a non-removable facial feature are templates calculated from images of individuals who do not possess the non-removable facial feature, which can reduce the degree of dissimilarity between two faces. In one example, the non-removable facial feature is facial hair.Thus, the one or more positive templates of a removable facial feature are templates calculated from images of people who have facial hair, and the one or more negative templates of a non-removable facial feature are templates calculated from images of people who do not have facial hair.

[0050] In some examples, the facial feature templates 20 can be a group of weighted weak classifiers, each consisting of a normalized similarity score between the face from which the facial feature template was generated and a facial image for which it is known whether the removable facial feature and / or the non-removable facial feature is present. In one example, the facial feature module 14 can assign a weight to each template of the facial feature templates 20 and determine, at least partially, based on a weighted sum of similarity scores between the facial template and the facial feature templates 20, whether the facial template from the authentication image contains the removable facial feature and / or the non-removable facial feature.In one example, the facial feature module 14 can use a learning technique known as AdaBoost (adaptive boosting) to iteratively select facial templates as weak classifiers from the templates of 60 removable facial features and from the templates of 62 non-removable facial features, and to weight them appropriately. The facial feature module 14 can adaptively influence the selection of training examples (e.g., of facial feature templates 20) in a way that improves the overall classifiers. More precisely, the training examples (e.g., the facial feature templates 20) are selected for each classifier (e.g.,(who possesses and does not possess the removable facial feature and who possesses and does not possess the non-removable facial feature) are weighted in such a way that the facial feature templates 20 that have been incorrectly classified by a given classifier are more likely to be selected for further training than facial feature templates that have been correctly classified.

[0051] As an example, AdaBoost is applied below to capture a removable facial feature (e.g., sunglasses) that can reduce the distinctiveness between two faces. In this example, this is referred to as... Xtraining={xi},i{1…N} a training set of N face templates designated for the presence of sunglasses on the face, and designated Xcross validation a cross-validation set of disjoint objects of M similarly labeled face templates. Furthermore, ω1 and ω2 denote the category identifiers, such as ω1 = sunglasses and ω2 = no sunglasses. Finally, h denotes i (x) a “weak classifier”, wherein hi(x)=g(pi×similarity(xi,x)) is. Here, p denotes i the polarity of the template x i in such a way that similarity (x) holds true. i , x) denotes the rating of the match for the face template x i and for the face case x given a face recognition device, and g() denotes a normalization function that applies to X training is calculated in such a way that hi(x) [−1,+1] applies.

[0052] The application of AdaBoost iteratively selects classifier h i (x) and weights them in such a way that the classification error over X trainingThe process is minimized. The process is repeated until the classification error exceeds X. cross-validation no longer decreases. The sunglasses classifier H(x) for a given face case x over ω1 and ω2 is given by H(x)=kΣ wk×hx(x) given. Here w denotes k the weighting for the classifier h k (x), which is assigned during training by the AdaBoost algorithm. Now, H(x) > 0 indicates the presence of sunglasses, while H(x) < 0 indicates the absence of sunglasses.

[0053] As discussed here, the computer device 10 may be in a secured state, and a user may wish to gain access to the computer device 10. In some examples, authentication can be granted using facial recognition (e.g., the facial recognition module 12). The camera 26 can capture an authentication image containing at least one user's face for a facial recognition process to unlock the computer device 10. The facial feature module 14 can receive the authentication image and compute an authentication face template for the face in the authentication image.The facial feature module 14 can analyze the authentication face template to determine whether an authentication image captured by the camera 26 contains a removable facial feature that can reduce a degree of dissimilarity between two faces, and / or a non-removable facial feature that can reduce a degree of dissimilarity between two faces.

[0054] For example, the facial feature module 14 can compare the authentication face template with templates of 60 removable facial features. Facial feature module 14 can determine multiple similarity scores of removable facial features between the authentication face template and one or more templates of 60 removable facial features. That is, for each comparison between the authentication face template and one or more positive templates of a removable facial feature and one or more negative templates of a removable facial feature, a similarity score of removable facial features can be determined. Facial feature module 14 can calculate a weighted sum of multiple similarity scores of removable facial features.As discussed here, facial feature module 14 can determine that the authentication image contains the removable facial feature if the weighted sum of the multiple similarity ratings of removable facial features is positive. Additionally, facial feature module 14 can determine that the authentication image does not contain the removable facial feature if the weighted sum of the multiple similarity ratings of removable facial features is negative.

[0055] The facial feature module 14 can send a message to the facial authentication module 16 indicating that the authentication face template of the authentication image contains a removable facial feature (e.g., sunglasses). In response to the message, the facial authentication module 16 can cause the output device 28 to issue a message to the user instructing them to remove the removable facial feature. The user can remove the removable facial feature, and the camera 26 can capture a second authentication image for facial recognition. The facial feature module 14 can calculate another authentication face template and determine whether the authentication image contains the removable facial feature. The facial feature module 14 can send a message to the facial authentication module 16 indicating that the authentication face template does not contain the removable facial feature.

[0056] The facial authentication module 16 can calculate a similarity score between the authentication face template (which, for example, does not possess the removable facial feature) and the registered face template and determine whether the first template exceeds the similarity score threshold. If the similarity score is higher than the similarity score threshold, the facial authentication module 16 can grant authentication for the computer device 10 by facial recognition and transition the computer device 10 from a secured state to an unlocked state. If the similarity score is lower than the similarity score threshold, the facial authentication module 16 can deny authentication for the computer device 10 by facial recognition and prevent the computer device 10 from transitioning from a secured state to an unlocked state.

[0057] Furthermore, the facial feature module 14 can compare the authentication face template with templates of 62 non-removable facial features. Facial feature module 14 can determine multiple similarity scores of non-removable facial features between the authentication face template and the templates of 62 non-removable facial features. That is, for each comparison between the authentication face template and one or more positive templates of a non-removable facial feature and one or more negative templates of a non-removable facial feature, facial feature module 14 can determine a similarity score of a non-removable facial feature. Facial feature module 14 can calculate a weighted sum of the multiple similarity scores of non-removable facial features.As discussed here, facial feature module 14 can determine that the authentication image contains the non-removable facial feature if the weighted sum of the multiple non-similarity ratings of removable facial features is positive. Furthermore, facial feature module 14 can determine that the authentication image does not contain the non-removable facial feature if the weighted sum of the multiple similarity ratings of non-removable facial features is negative.

[0058] The facial feature module 14 can send a message to the facial authentication module 16 indicating that the facial template in the authentication image contains a non-removable facial feature (e.g., facial hair). In response to this message, the facial authentication module 16 can increase the similarity assessment threshold to a predefined threshold. The facial recognition module 16 can then calculate a similarity score between the user's facial template in the authentication image and one or more registered templates of authorized users (e.g., the user). The facial authentication module 16 can determine whether the similarity score is higher than the predefined threshold.If the similarity score is higher than the set similarity score, the facial authentication module 16 can grant authentication for the computer device 10 by facial recognition and transition the computer device 10 from a secured state to an unlocked state. If the similarity score is lower than the set similarity score, the facial authentication module 16 can deny authentication for the computer device 10 by facial recognition and prevent the computer device 10 from transitioning from a secured state to an unlocked state.

[0059] In some examples, the facial feature module 14 can determine whether the authentication face template contains a removable facial feature and / or a non-removable facial feature before the facial authentication module 16 begins to determine whether the authentication image matches one of the templates registered by the user. For example, the facial authentication module 16 can compare the authentication face template with user-registered templates 22 if the facial authentication module 16 receives a message from the facial feature module 14 indicating that the authentication face template does not contain a removable facial feature or a non-removable facial feature that can reduce the distinctiveness between two faces.For example, the facial feature module 14 can prevent the facial authentication module 16 from analyzing the authentication face template to determine whether the user is an authenticated user if a removable and / or non-removable facial feature, which can reduce the distinctiveness between two faces, is detected. In some examples, preventing the facial authentication module 16 from performing facial recognition techniques can save power and extend battery life.

[0060] In further examples, the facial feature module 14 can analyze the authentication template for the removable facial feature and for the non-removable facial feature, and the facial authentication module 16 can essentially simultaneously compare the authentication facial template with user-registered templates 22. If the facial feature module 14 determines that the authentication facial template contains the removable facial feature and / or the non-removable facial feature, the facial feature module 14 can send a message to the facial authentication module 16 indicating that the facial authentication module should stop the comparison.

[0061] In other examples, the facial authentication module 16 can determine that the authentication template matches one or more user-registered templates before the facial feature module 16 determines whether the authentication template contains the removable facial feature and / or the non-removable facial feature. If the facial authentication module 16 determines that the authentication template matches one or more user-registered templates before the facial feature module 14 determines whether the authentication template contains the removable facial feature and / or the non-removable facial feature... If the authentication face template does not contain the removable or non-removable facial feature, the facial authentication module 16 can wait to transition the computer device 10 from a secured state to an unlocked state until it receives a message from the facial feature module 14 indicating that the authentication face template does not contain the removable or non-removable facial feature. When the facial authentication module 14 receives the message indicating that the authentication face template does not contain the removable or non-removable facial feature, it can grant access to the computer device 10.

[0062] In another example, the facial feature module 14 can send a message indicating that the authentication face template contains the removable facial feature and / or the non-removable facial feature after the facial authentication module 16 has determined that the authentication face template matches the registered user template. In this case, the facial authentication module 16 can deny access to the computer device 10 or perform another facial recognition with the set similarity assessment threshold, even though the authentication face template was a match.For example, if the facial feature module 14 detects a removable facial feature after the facial authentication module 16 has determined that the authentication template matches one or more templates registered by the user, the facial authentication module 16 can deny access and display a message to the user instructing them to remove the removable facial feature. In some examples, the message to the user may indicate that another authentication image should be captured to continue with facial recognition. If the facial feature module 14 detects a non-removable facial feature after the facial authentication module 16 has determined that the authentication template matches one or more templates registered by the user, the facial authentication module 16 can reduce the security (e.g.,by increasing the similarity assessment threshold to a set similarity assessment threshold) and re-executing facial recognition based on the increased security (e.g., to determine whether the authentication face template matches the image registered by the user). That is, the facial authentication module 16 can determine whether the similarity assessment is higher than the set similarity assessment threshold.

[0063] Fig. Figure 3 is a conceptual diagram illustrating the behavior of the computer device 10 after a user 70 has attempted to unlock the computer device 10 using facial recognition technology. The user 70 can authenticate themselves to the computer device 10 using facial recognition. The GUI 76 can display graphical information relating to the authentication of a user to the computer device 10 using facial recognition in accordance with techniques of this disclosure. The GUI 76 can include one or more GUI elements, such as a security indicator 78 and a recording icon 80. The computer device 10 can be configured to operate in a "secured" mode, as indicated by the security indicator 78.

[0064] In the specific example from Fig. 3. User 70 may have taken an authentication image (e.g., an image used for facial recognition to transition the computer device 10 from a secured state to an unlocked state) using an input device such as a camera 82. Because User 70 is wearing sunglasses (e.g., a removable facial feature that can reduce the degree of distinctiveness between two faces), User 70 cannot authenticate by facial recognition.

[0065] In this example, it can be assumed that user 70 is an authorized user (i.e., that at least one user registration template stored by computer device 10 was calculated from an image containing user 70's face). After computer device 10 detects that user 70 has a removable facial feature that can reduce the degree of distinctiveness between two faces (e.g., sunglasses), computer device 10 can issue a message 74. The message 74 can indicate that user 70 has not gained access to computer device 10. In this example, the message 74 can indicate that the reason computer device 10 did not authenticate user 70 is that the sunglasses were detected. In additional examples, the message 74 can include an action that requires the user to, for example,instructs the user to remove the sunglasses to continue the facial recognition process. In one example, the user can remove the removable facial feature (70) and proceed by selecting the capture icon (80) to take another authentication image for facial recognition.

[0066] As discussed here, certain removable facial features (e.g., sunglasses) can reduce the distinctiveness between two faces and cause an otherwise unauthorized user to gain access to computer device 10 if the unauthorized user possesses the removable facial feature. Computer device 10 can detect when a user attempting to access computer device 10 has the removable facial feature and can issue a message indicating that the authentication process failed because sunglasses were detected. The message can instruct the user to remove the sunglasses to proceed. In one example, computer device 10 can begin by taking an authentication image of a user attempting to access computer device 10.As discussed here, the computer device 10 can use one or more facial recognition programs executed on the computer device 10 to check the match between an authentication face template calculated from the authentication image and one or more templates registered by the user. To reduce the occurrence of errors by the facial recognition programs, the computer device 10 can first analyze the authentication template for one or more removable facial features that could contribute to such errors. For example, the computer device 10 can analyze the authentication template for one or more removable facial features that could reduce the dissimilarity between two faces.

[0067] Fig. Figure 4 is a conceptual diagram illustrating the behavior of the computer device 10 after a user 90 has attempted to unlock the computer device 10 using facial recognition technology. The user 70 can authenticate themselves to the computer device 10 using facial recognition. The GUI 76 can display graphical information relating to the authentication of a user to the computer device 10 using facial recognition in accordance with the techniques of this disclosure. The GUI 76 can include one or more GUI elements, such as a security indicator 78 and a recording icon 80. The computer device 10 can be configured to operate in a "secured" mode, as indicated by the security indicator 78.

[0068] In the specific example from Fig. 4. The user 90 may have captured an authentication image through an input device such as a camera 82. As in Fig. As shown in Figure 4, the facial feature module 14 has detected a non-removable facial feature that can reduce the difference between two faces (e.g., facial hair) and sent a message to the facial authentication module 16 to stop the facial recognition process.

[0069] In this example, user 90 can be an authorized user (i.e., at least one user registration template stored by computer device 10 was calculated from an image containing user 90's face). After computer device 10 detects that user 90 has a non-removable facial feature that can reduce the degree of distinctiveness between two faces (e.g., facial hair), computer device 10 can issue a message 94. The message 94 can indicate that the facial recognition process has been stopped because facial hair has been detected. Furthermore, the message 94 can indicate that security has been increased.As discussed here, the facial authentication module 16 can confirm a user's authentication if a similarity score between the template calculated from the authentication image and one or more of the registered user templates exceeds a similarity score threshold. When detecting facial hair, the computer device 10 can adjust the similarity score accordingly. Increase a set similarity assessment threshold. Additionally, the notification can provide options (e.g., "Yes" and "No") in case the user wishes to proceed with facial recognition with the increased security.

[0070] In some examples, the facial feature module 14 cannot cause the computer device 10 to issue a message 94 to the user 90 when the non-removable facial feature is captured. In this case, the computer device 10 can automatically increase security. Access to the computer device 10 depends on whether the similarity score between the authentication template and the one or more templates registered by the user is higher than the set similarity score threshold.

[0071] As discussed here, certain non-removable facial features (e.g., facial hair) can reduce the distinctiveness between two faces and allow an otherwise unauthorized user to gain access to the computer device 10 if the unauthorized user possesses the non-removable facial feature. The computer device 10 can detect when a user attempting to gain access to the computer device 10 possesses the non-removable facial feature and can increase the similarity assessment threshold to a preset similarity rating. The facial feature module 14 can cause or prevent the computer device 10 from issuing a message to the user indicating that security has been increased.As discussed here, the computer device 10 can use the one or more facial recognition programs running on the computer device 10 to check whether an authentication template calculated from the authentication image matches one or more stored templates registered by the user. To reduce the occurrence of errors by the facial recognition programs, the computer device 10 can first analyze the authentication template for one or more non-removable facial features that could contribute to such errors. For example, the computer device 10 can analyze the authentication template for one or more non-removable facial features that could reduce the dissimilarity between two faces.

[0072] Fig. Figure 5 is a flowchart representing an exemplary process that can be executed by a computer device to determine whether an image for facial recognition contains a removable facial feature and / or a non-removable facial feature that can reduce the degree of distinctiveness between two faces. Process 500 can be executed by any computer device described in this disclosure. For the sake of clarity, Process 500 is here referred to in relation to computer device 10. Fig. 1 in the context of the in relation to Fig. 24 described analysis methods.

[0073] Process 500 can begin when the computer device 10 captures an image (502). In many cases, the computer device 10 can use an image capture device such as a camera 26. The computer device 10 can compute a face template from the image (504). Although the captured image and face template can serve multiple purposes, Process 500 is described only for the sake of clarity in relation to an authentication image and an authentication face template. In one example, the face feature module 14 and / or the face authentication module 16 can compute the authentication face template from the authentication face image.

[0074] The computer device 10 can analyze the face template for a removable facial feature and / or a non-removable facial feature that can reduce the degree of distinctiveness between two faces (506). As discussed here, the computer device 10 can store multiple facial feature templates 20. The facial feature module 14 can compare the face template with the multiple facial feature templates 20 stored in the facial recognition database 18 to determine whether the face template contains a removable and / or a non-removable facial feature that can reduce the degree of distinctiveness between two users.

[0075] The facial feature templates 20 can contain a group of templates 60 of removable facial features and a group of templates 62 of non-removable facial features. The group of templates 60 of removable facial features can contain one or more positive templates of a removable facial feature and one or more negative templates of a removable facial feature, wherein the positive templates of a removable facial feature are templates calculated from images of people who possess the removable facial feature, and the negative templates of a removable facial feature are templates calculated from images of people who do not possess the removable facial feature.Furthermore, the group of templates 62 non-removable facial features can contain one or more positive templates of a removable facial feature and one or more negative templates of a removable facial feature, wherein the positive templates of a removable facial feature are templates calculated from images of persons possessing the non-removable facial feature and the negative templates of a removable facial feature are templates calculated from images of persons not possessing the non-removable facial feature.

[0076] Process 500 can determine whether the facial feature contains a removable facial feature that can reduce the dissimilarity between two faces (508). In an example, feature module 14 can determine whether the facial template contains the removable facial feature (508). For example, as discussed here, facial feature module 14 can compare the facial template with the set of templates 60 removable facial features. Facial feature module 14 can determine multiple similarity ratings of removable facial features between the facial template and at least one of the one or more positive templates of a removable facial feature and at least one of the negative templates of a removable facial feature. Facial feature module 14 can compute a weighted sum of the multiple similarity ratings of removable facial features.In some examples, the face template includes the removable facial feature if the weighted sum of the multiple similarity ratings of removable facial features is positive, and the face template does not include the removable facial feature if the weighted sum of the multiple similarity ratings of removable facial features is negative.

[0077] If the face template contains a removable facial feature (the "yes" branch from 508), process 500 can issue a message to the user instructing them to remove the removable facial feature (510). For example, the face feature module 14 can send a message to the face authentication module 16 indicating that the removable facial feature has been detected in the face template. In response to the message, the face authentication module 16 can cause the computer device 10 to issue a message to the user instructing them to remove the removable facial feature. A user can take another picture for the face recognition process. For example, process 500 can include restarting the process by taking a picture (502) when the face authentication module 16 issues a message to the user instructing them to remove the removable facial feature.

[0078] If the face template does not contain the removable facial feature (“No” branch from 508), process 500 can determine whether the face template contains the non-removable facial feature (512). For example, as discussed here, the face feature module 14 can compare the face template with the group of templates 62 of non-removable facial features. The face feature module 14 can determine multiple similarity ratings of non-removable facial features between the face template and at least one of the one or more positive templates of a non-removable facial feature and at least one of the one or more negative templates of a non-removable facial feature. The face feature module 14 can compute a weighted sum of the multiple similarity ratings of removable facial features.In some examples, the face template includes the non-removable facial feature if the weighted sum of the multiple similarity ratings of non-removable facial features is positive, and the face template does not include the non-removable facial feature if the weighted sum of the multiple similarity ratings of non-removable facial features is negative.

[0079] If the face template contains the non-removable facial feature (the "yes" branch from 512), process 500 can set a first similarity assessment threshold to a second similarity assessment threshold (514). For example, the face feature module 14 can send a message to the face authentication module 16 indicating that the non-removable facial feature 16 has been captured in the face template. In response to the message, the face authentication module 16 can increase the similarity assessment threshold to a set similarity assessment threshold.

[0080] If the face template does not contain the non-removable face feature (“No” branch from 512), process 500 can compute a similarity score between the face template and one or more registered face templates (512). For example, the face feature module 14 can send a message to the face authentication module 16 indicating that the face template does not contain the removable face feature or the non-removable face feature that can reduce the dissimilarity between two faces.

[0081] In response to the message, the facial authentication module 16 can proceed with the facial recognition process and compute a similarity score between the face template and the one or more registered face templates. As discussed here, the user's registered template(s) are a computed template of one or more previously captured images of the user. Process 500 can determine whether the similarity score is higher than the first similarity score threshold (518). If the similarity score is higher than the first similarity score threshold (the "yes" branch of 518), Process 500 can grant authentication (520). For example, the facial authentication module 16 can determine that the similarity score is higher than the first similarity score threshold and authenticate the user attempting to access the computer device 10.For example, granting authentication may involve transitioning computer device 10 from a secured state to an unsecured state.

[0082] If the similarity score is less than the first similarity score threshold (the "no" branch from 518), process 500 can deny authentication (522). For example, facial authentication module 16 might determine that the similarity score is less than the first similarity score threshold and not authenticate the user attempting to access computer device 10. Denying authentication might include, for example, preventing computer device 10 from transitioning from a secured state to an unsecured state.

[0083] In some examples, the face recognition module 12 analyzes the face template to determine whether the face image contains a removable facial feature and / or a non-removable facial feature that can reduce the degree of distinctiveness between two faces. Additionally, face recognition module 12 calculates the similarity score between the face template of the first image and the registered template of authorized users. As discussed here, analyzing the first face template for facial features that can reduce the degree of distinctiveness and analyzing the first face template for face recognition authentication can be performed using the same face recognition algorithm.For example, the facial recognition algorithm, as discussed here, is trained to detect specific facial features that can reduce the degree of distinctiveness between users. As discussed here, in some examples, the removable facial feature that can reduce the degree of distinctiveness between two faces is sunglasses, and the non-removable facial feature that can reduce the degree of distinctiveness between two faces is facial hair.

[0084] Fig. Figure 6 is a flowchart representing an exemplary process that can be executed by a computer device when a removable facial feature has been captured in the image for facial recognition, which can reduce the degree of distinctiveness between two faces. Process 600 can be executed by any computer device described in this disclosure. For the sake of clarity, process 600 is here referred to in relation to computer device 10. Fig. 1 in the context of the in relation to Fig. 24 described analysis methods.

[0085] Process 600 can begin when the computer device 10 detects that the face template contains the removable facial feature and sends a message to the user to remove the removable facial feature (e.g., step 510 in the [document / section]). Fig. Process 500 (as shown in Figure 5) is output. In some examples, Process 600 can begin by capturing another image (e.g., a second image) for face recognition (602). In many cases, the computer device 10 can use an image capture device such as a camera 26. The computer device 10 can compute another face template (e.g., a second face template) from the second image (604). Although the second image and the second face template can serve multiple purposes, Process 600 is described only for the sake of clarity in terms of a second authentication image and a second authentication face template. In one example, the face feature module 14 and / or the face authentication module 16 can compute the second authentication face template from the second authentication face image.

[0086] Furthermore, process 600 can calculate a similarity score between the second face template and the registered face template (606). For example, the face feature module 14 can send a message to the face authentication module 16 indicating that the face template does not contain the removable face feature or the non-removable face feature that can reduce the distinctiveness between two faces. Process 600 can include analyzing the second face template for a removable face feature and / or a non-removable face feature that reduces a degree of distinctiveness between two faces (e.g., step (506) from process 500).

[0087] Fig. Figure 7 is a flowchart representing an exemplary process that can be executed by a computer device when a non-removable facial feature has been captured in the image for facial recognition, which can reduce the degree of distinctiveness between two faces. Process 700 can be executed by any computer device described in this disclosure. For the sake of clarity, process 700 is here expressed in relation to computer device 10. Fig. 1 in the context of the in relation to Fig. 24 described analysis methods.

[0088] Process 700 can begin when computer device 10 detects that the face template contains the non-removable facial feature and sets a first similarity assessment threshold to a second similarity assessment threshold (e.g., step 514 in the Fig.(Process 500, as depicted in Figure 5). For example, the facial feature module 14 can send a message to the facial authentication module 16 indicating that the non-removable facial feature 16 has been captured in the facial template. In response to the message, the facial authentication module 16 can adjust the first similarity assessment threshold to a second similarity assessment threshold, where the second similarity assessment threshold is higher than the first similarity assessment threshold.

[0089] In some examples, process 700 can calculate a similarity score between the face template and the registered face template (702). Process 700 can determine whether the similarity score is higher than the second similarity score threshold (704). If the similarity threshold is higher than the second similarity score threshold (the "yes" branch from 704), process 700 can include granting facial authentication (706). For example, facial authentication module 16 can determine that the similarity score is higher than the second similarity score threshold and authenticate the user attempting to access computer device 10. Upon granting facial authentication, process 700 can transition computer device 10 from a secure state to an unsecured state (708).

[0090] If the similarity score is less than the second similarity score threshold (the "no" branch from 708), process 700 can deny authentication (710). The facial authentication module 16 can determine that the similarity score is less than the set similarity score threshold and may not authenticate the user attempting to access the computer device 10. Process 700 can prevent the computer device 10 from transitioning from a secured state to an unsecured state (712).

[0091] In some examples, process 700 can determine whether an authentication session is complete. For example, determining that an authentication session is complete can include determining whether facial recognition authentication was denied and / or granted. In one example, facial authentication module 16 can determine whether the authentication session has ended based on whether facial recognition access was granted to the authentication template. Furthermore, process 700 can include resetting the second similarity assessment threshold to the first similarity assessment threshold in response to the determination that the authentication session is complete. For example, facial authentication module 16 can reduce the second similarity assessment threshold back to the first similarity assessment threshold when the authentication session is complete.

[0092] An exemplary method comprises capturing an image containing at least one user's face with a computer camera, calculating a face template of the user's face in the image, and analyzing the face template to determine whether the face contains a removable facial feature that reduces the degree of dissimilarity between two faces, and / or a non-removable facial feature that reduces the degree of dissimilarity between two faces. If the face contains the removable facial feature, the method further comprises issuing a message to the user instructing them to remove the removable facial feature. If the face contains the non-removable facial feature, the method further comprises setting a first similarity assessment threshold to a second similarity assessment threshold.

[0093] The techniques described herein may be implemented, at least partially, in hardware, software, firmware, or any combination thereof. For example, various aspects of the described embodiments may be implemented within one or more processors, including one or more microprocessors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), free-programmable logic arrays (FPGAs), or any other equivalent integrated or discrete logic circuit arrangement, as well as any combination of such components. The term "processor" or "processing circuit arrangement" may generally refer to any of the foregoing logic circuit arrangements, alone or together with another logic circuit arrangement, or to any other equivalent circuit arrangement.A control unit containing hardware can also execute one or more of the techniques of this disclosure.

[0094] This hardware, software, and firmware can be implemented within the same device or within separate devices to support the various techniques described herein. Furthermore, any of the units, modules, or components described herein can be implemented together or separately as a discrete but interoperable logic device. The representation of various features as modules or units is intended to highlight different functional aspects and does not necessarily imply that these modules or units are implemented by separate hardware, firmware, or software components. Rather, the functionality associated with one or more modules or units can be implemented by separate hardware, firmware, or software components or integrated within shared or separate hardware, firmware, or software components.

[0095] The techniques described herein can also be embodied or encoded in a manufactured article including a computer-readable storage medium encoded with instructions. The instructions embedded or encoded in a manufactured article containing an encoded computer-readable storage medium can cause one or more programmable processors or other processors to implement one or more of the techniques described herein, such as when instructions contained or encoded in the computer-readable storage medium are executed by the one or more processors.Computer-readable storage media can include read / write memory (RAM), read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electronically erasable programmable read-only memory (EEPROM), flash memory, a hard disk, a compact disc ROM (CD-ROM), a floppy disk, a cassette, magnetic media, optical media, or other computer-readable media. In some examples, a manufactured item may include one or more computer-readable storage media.

[0096] In some examples, the computer-readable storage media may include non-volatile media. The term "non-volatile" can indicate that the storage medium is concrete and not embodied in a carrier wave or propagated signal. In certain examples, a non-volatile storage medium may store data that can change over time (e.g., in RAM or a cache).

Claims

[1] Procedure that includes: Storing a set of templates (60) of removable facial features and a set of templates (62) of non-removable facial features; Weights of each of the plurality of templates (60) of removable facial features and the plurality of templates (62) of non-removable facial features such that the facial feature templates that have been incorrectly classified by a given classifier are more likely to be selected for further training than facial feature templates that have been correctly classified; Capturing an image containing at least one user's face (30) by a computer device (10); Calculating a face template of the user's face (30) in the image using the computer device (10); Determine whether the face contains a removable facial feature that reduces a degree of dissimilarity between two faces, by the computer device (10) and at least partially on the basis of a weighted sum of similarity assessments between the face template and each of the plurality of templates (60) of removable facial features; Determine whether the face contains a non-removable facial feature that reduces a degree of dissimilarity between two faces, by the computer device (10) and at least partially on the basis of a weighted sum of similarity assessments between the face template and each of the plurality of templates (62) of non-removable facial features; wherein the procedure, if the face contains the removable facial feature, comprises: issuing a message to the user (30) that he should remove the removable facial feature, by means of the computer device (10); and wherein the procedure, if the face contains the non-removable facial feature, comprises: setting a first similarity assessment threshold to a second similarity assessment threshold by the computer device (10). [2] Method according to claim 1, wherein the group of templates (60) of removable facial features comprises one or more positive templates of a removable facial feature and one or more negative templates of a removable facial feature, wherein the positive templates of a removable facial feature are images of persons who possess the removable facial feature, and wherein the negative templates of a removable facial feature are images of persons who do not possess the removable facial feature. [3] The method of claim 2, further comprising: Comparing the face template with the group of templates (60) of removable facial features by the computer device (10); The computer device (10) includes determining multiple similarity assessments of removable facial features between the facial template and one or more of the positive templates of a removable facial feature and one or more of the negative templates of a removable facial feature; Calculating the weighted sum of the multiple similarity ratings of removable facial features by the computer device (10), where it is determined that the face contains the removable facial feature if the weighted sum of the multiple similarity ratings of removable facial features is positive, and where it is determined that the face does not contain the removable facial feature if the weighted sum of the multiple similarity ratings of removable facial features is negative. [4] Method according to any one of claims 1 to 3, wherein the image is a first image and wherein the face template is a first face template and wherein the method, if the face contains the removable facial feature, further comprises: Capturing a second image containing the user's face (30) by the computer device (10); Calculating a second face template of the user's face (30) in the second image using the computer device (10); Calculating a similarity score between the user's second face template (30) in the second image and a registered template (22) of the user (30); and Determine whether the similarity assessment is above the first similarity assessment threshold using the computer device (10). [5] Method according to claim 4, wherein the registered template (22) of the user (30) is calculated from a previously recorded image of the user (30). [6] Method according to claim 4 or 5, wherein authentication is granted by facial recognition when it is determined that the similarity assessment is higher than the first similarity assessment threshold. [7] Method according to any one of claims 4 to 6, wherein authentication by facial recognition is denied when it is determined that the similarity assessment is less than the first similarity assessment threshold. [8] Method according to any one of claims 1 to 7, wherein the group of templates (62) of non-removable facial features comprises one or more positive templates of a non-removable facial feature and one or more negative templates of a non-removable facial feature, wherein the positive templates of a non-removable facial feature are images of persons who possess the non-removable facial feature, and wherein the negative templates of a removable facial feature are images of persons who do not possess the non-removable facial feature. [9] The method of claim 8, further comprising: Comparing the face template with the group of templates (62) of non-removable facial features by the computer device (10); Determining multiple similarity assessments of non-removable facial features between the facial template and one or more of the positive templates of a non-removable facial feature and each of the negative templates of a non-removable facial feature by the computer device (10); and Calculating a weighted sum of the multiple similarity ratings of non-removable facial features by the computer device (10), where it is determined that the face contains the non-removable facial feature if the weighted sum of the multiple similarity ratings of non-removable facial features is positive, and where it is determined that the face does not contain the non-removable facial feature if the weighted sum of the multiple similarity ratings of non-removable facial features is negative. [10] Method according to any one of claims 1 to 9, wherein the method, if the face contains the non-removable facial feature, further comprises: Calculating a similarity score between the user's face template (30) in the image and a registered template (22) of the user (30); and Determine whether the similarity assessment is above the second similarity assessment threshold using the computer device (10). [11] The method of claim 10, wherein the method further comprises, if the similarity assessment is above the second similarity assessment threshold: Granting authentication through facial recognition; and Transferring the computer device (10) from a secured state to an unsecured state, and wherein, if the similarity assessment is below the second similarity assessment threshold, the procedure further comprises: Denying authentication via facial recognition; and Prevent the computer device (10) from transitioning from a secured state to an unlocked state. [12] The method of claim 10 or 11, further comprising: Determine when an authentication session is complete, by the computer device (10); and Resetting the second similarity assessment threshold to the first similarity assessment threshold in response to the determination that the authentication session is complete. [13] The method of claim 12, wherein determining when the authentication session is complete further comprises: Determine whether authentication by facial recognition was denied and / or granted by the computer device (10). [14] Method according to any one of claims 1 to 13, wherein the method, if the first face template does not contain the removable facial feature and the non-removable facial feature, further comprises: Calculating a similarity score between the user's face template (30) in the image and a registered template (22) of the user (30) by the computer device (10); and Determine whether the similarity assessment is above the first similarity assessment threshold using the computer device (10), where the procedure further includes, if the similarity assessment is above the first similarity assessment threshold: Granting authentication through facial recognition; and Transferring the computer device (10) from a secured state to an unsecured state; and wherein, if the similarity assessment is below the first similarity assessment threshold, the procedure further comprises: Denying authentication via facial recognition; and Prevent the computer device (10) from transitioning from a secured state to an unlocked state. [15] Method according to any one of claims 1 to 14, wherein the analysis of the first face template to determine whether the face contains a removable facial feature that reduces a degree of dissimilarity between two faces and / or a non-removable facial feature that reduces a degree of dissimilarity between two faces comprises the analysis of the first face template using a face recognition module, and wherein the calculation of the similarity assessment between the first face template of the user (30) in the first image and the registered template (22) of the user (30) comprises the calculation by the face recognition module. [16] Method according to any one of claims 1 to 15, wherein the removable facial feature comprises sunglasses. [17] Method according to any one of claims 1 to 16, wherein the non-removable facial feature comprises facial hair. [18] Method according to any one of claims 1 to 17, wherein the second similarity assessment threshold is greater than the first similarity assessment threshold. [19] A computer-readable storage device that stores instructions to cause at least one processor (50) of a computer device (10) to perform operations comprising: Storing a set of templates (60) of removable facial features and a set of templates (62) of non-removable facial features; Weights of each of the plurality of templates (60) of removable facial features and the plurality of templates (62) of non-removable facial features such that the facial feature templates that have been incorrectly classified by a given classifier are more likely to be selected for further training than facial feature templates that have been correctly classified; Capturing an image containing at least one user's face (30) by a camera (26) of the computer device (10); Calculating a face template of the user's face (30) in the image; and Determine whether the face contains a removable facial feature that reduces a degree of dissimilarity between two faces, by the computer device (10) and at least partially on the basis of a weighted sum of similarity assessments between the face template and each of the plurality of templates (60) of removable facial features; Determine whether the face contains a non-removable facial feature that reduces a degree of dissimilarity between two faces, by the computer device (10) and at least partially on the basis of a weighted sum of similarity assessments between the face template and each of the plurality of templates (62) of non-removable facial features; Outputting a message to the user (30) that he should remove the removable facial feature, in response to the determination that the face contains the removable facial feature, through a graphical user interface (32) of the computer device (10); and Adjusting a first similarity assessment threshold to a second similarity assessment threshold in response to the determination that the face contains the non-removable facial feature. [20] Computer device comprising: a memory (52) for storing a group of templates (60) of removable facial features and a group of templates (62) of non-removable facial features; at least one processor (50); at least one camera (26) that is operable by the at least one processor (50) for capturing a first image containing at least one first face of a user (30); and at least one output device (28), where at least one processor (50) is configured to perform the following operations: Weights of each of the plurality of templates (60) of removable facial features and the plurality of templates (62) of non-removable facial features such that the facial feature templates that have been incorrectly classified by a given classifier are more likely to be selected for further training than facial feature templates that have been correctly classified; Capturing an image containing at least one user's face (30) by the camera (26) of the computer device (10); Calculating a face template of the user's face (30) in the image; Determine whether the face contains a removable facial feature that reduces a degree of dissimilarity between two faces, by the computer device (10) and at least partially on the basis of a weighted sum of similarity assessments between the face template and each of the plurality of templates (60) of removable facial features; and Determine whether the face contains a non-removable facial feature that reduces a degree of dissimilarity between two faces, by the computer device (10) and at least partially on the basis of a weighted sum of similarity assessments between the face template and each of the plurality of templates (62) of non-removable facial features; wherein at least one output device (28) is configured to output a message to the user (30) that he should remove the removable facial feature when the face contains the removable facial feature, and wherein the at least one processor (50) sets a first similarity assessment threshold to a second similarity assessment threshold when the face contains the non-removable facial feature.

Images