Machine learning model for denoising biometric signals collected using mobile biometric sensors

Abstract

Training data is generated for training a machine learning biometric denoising model. Generating the training data includes obtaining information describing multiple clean biometric signals measured using a corresponding medical device. For each clean biometric signal, noise is sampled from a noise source among multiple candidate noise sources. This noise is applied to at least a portion of the clean biometric signal. The training data is processed using the machine learning biometric denoising model to obtain multiple denoised biometric signals. The machine learning biometric denoising model is trained based on a loss function that evaluates the differences between the multiple denoised biometric signals and the multiple clean biometric signals.

Description

[0001] Priority Statement

[0002] This application is based on and claims priority to U.S. Provisional Application No. 63 / 573,239, filed April 2, 2024. The application claims priority and benefits from each of these applications, and all such applications are incorporated herein by reference in their entirety. Technical Field

[0003] This disclosure generally relates to denoising biometric signals. More specifically, this disclosure relates to machine learning models for removing noise from biometric signals collected using mobile biometric sensors. Background Technology

[0004] In recent years, biometric sensors have become an increasingly important competitive area for developers of mobile computing devices (such as smartphones, smartwatches, and wearable devices). Some mobile computing devices are able to collect biometric signals that were previously collected using only specially customized medical devices. For example, blood oxygen levels were historically measured using specially customized medical devices, such as pulse oximeters attached to a user's finger. However, some recently developed mobile computing devices include sensors that can periodically measure blood oxygen levels while the device is worn by the user. For example, wearable smartwatches may include a pulse oximeter sensor on the underside of the watch, allowing the sensor to directly contact the user's wrist. In this way, biometric information can be periodically collected from the user to monitor health status and provide statistical information, etc. Summary of the Invention

[0005] Aspects and advantages of embodiments of this disclosure will be set forth in part in the description which follows, or may be learned from the description or by practice of the embodiments.

[0006] One example aspect of this disclosure relates to a computer-implemented method. The method includes: generating training data by a computing system including one or more processor devices for training a machine learning biometric denoising model, wherein generating the training data includes information obtained by the computing system describing a plurality of clean biometric signals measured using a corresponding medical device. The method includes: for each of the plurality of clean biometric signals, sampling noise from one or more noise sources from a plurality of candidate noise sources by the computing system. The method includes: for each of the plurality of clean biometric signals, applying noise to at least a portion of the clean biometric signal by the computing system. The method includes: processing the training data with the machine learning biometric denoising model by the computing system to obtain a plurality of denoised biometric signals. The method includes: training the machine learning biometric denoising model by the computing system based on a loss function that evaluates the differences between the plurality of denoised biometric signals and the plurality of clean biometric signals.

[0007] Another example aspect of this disclosure relates to a user computing device including one or more processor devices and one or more tangible non-transitory computer-readable media storing computer-readable instructions that, when executed by the one or more processor devices, cause the user computing device to perform operations. These operations include: acquiring a noisy biometric signal via a mobile biometric measurement device associated with the user computing device, wherein the noisy biometric signal includes a first noise amount. These operations include: causing the noisy biometric signal to be processed by a machine learning biometric denoising model to obtain a denoised biometric signal including a second noise amount less than the first noise amount, wherein the machine learning biometric denoising model is trained based on a loss function that evaluates the difference between a plurality of noisy biometric training signals and a plurality of clean biometric ground truth signals, and wherein the plurality of noisy biometric training signals are generated by applying noise from a randomly selected noise source to a randomly selected portion of a clean biometric signal. These operations include: providing biometric measurements to a user of the user computing device, wherein the biometric measurements are at least partially based on the denoised biometric signal.

[0008] Another example aspect of this disclosure relates to one or more tangible non-transitory computer-readable media storing computer-readable instructions that, when executed by one or more processor devices, cause the one or more processor devices to perform operations. These operations include: generating training data for training a machine learning biometric denoising model, wherein generating the training data includes obtaining information describing a plurality of clean biometric signals measured using a corresponding medical device. These operations include: for each of the plurality of clean biometric signals, selecting at least a portion of the clean biometric signal. These operations include: for each of the plurality of clean biometric signals, sampling noise from one or more noise sources from a plurality of candidate noise sources. These operations include: for each of the plurality of clean biometric signals, applying noise to at least that portion of the clean biometric signal. These operations include: processing the training data with a machine learning biometric denoising model to obtain a plurality of denoised biometric signals. These operations include: training the machine learning biometric denoising model based on a loss function that evaluates the differences between the plurality of denoised biometric signals and the plurality of clean biometric signals.

[0009] Other aspects of this disclosure relate to various systems, devices, non-transitory computer-readable media, user interfaces, and electronic devices.

[0010] These and other features, aspects, and advantages of the various embodiments of this disclosure will be better understood with reference to the following description and the appended claims. The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate exemplary embodiments of the disclosure and, together with the description, serve to explain the relevant principles. Attached Figure Description

[0011] Referring to the accompanying drawings, a detailed discussion of embodiments is set forth in this specification for those skilled in the art, in which: Figure 1A , Figure 1B and Figure 1C Each of the above shows a perspective view of an example non-limiting computing device (e.g., a wearable computing device) according to one or more example embodiments of the present disclosure.

[0012] Figure 2 This is a block diagram of an environment suitable for implementing a machine learning model for denoising biometric signals collected using mobile biometric sensors, based on some implementations of this disclosure.

[0013] Figure 3 It is based on some implementations of this disclosure for use Figure 2 A block diagram of the training data generator for the model training module.

[0014] Figure 4 A block diagram of an example non-limiting wearable device according to one or more example embodiments of the present disclosure is shown.

[0015] Figure 5 An illustration is shown of an example non-limiting user evaluation management system according to one or more example embodiments of the present disclosure.

[0016] Figure 6 A flowchart is depicted illustrating an example method for training a machine learning model for denoising biometric signals collected using a mobile biometric sensor, according to an example embodiment of the present disclosure.

[0017] Figure 7 A flowchart is depicted illustrating an example method, according to an example embodiment of the present disclosure, utilizing a machine learning model for denoising biometric signals collected using a mobile biometric sensor.

[0018] The repeated reference numerals across multiple figures are intended to identify the same features in various implementations. Detailed Implementation

[0019] Overview

[0020] Generally, this disclosure relates to denoising biometric signals. More specifically, this disclosure relates to removing noise from biometric signals collected using mobile biometric sensors. In recent years, accurate biometric sensors have become an increasingly competitive area in the development of wearable computing devices. Some wearable computing devices are capable of collecting biometric signals that were previously collected only using specially designed medical devices. For example, blood oxygen levels have historically been measured using specially designed medical devices, such as pulse oximeters attached to a user's finger. However, recent wearable computing devices include sensors capable of measuring blood oxygen levels when the device is worn by a user. For example, a wearable smartwatch device may include a pulse oximeter sensor on the underside of the watch, allowing the sensor to directly contact the user's wrist.

[0021] In some cases, conventional medical sensors and / or signal collection technologies can be incorporated into wearable computing devices without substantial modification. Continuing with the previous example, both a wearable smartwatch attached to a user's wrist and a pulse oximeter attached to a user's finger measure blood oxygenation by projecting infrared light onto the user's skin. However, other types of medical sensors must be substantially modified before they can be incorporated into certain wearable computing devices. For example, in medical settings, ECG signals are collected by placing a large number of electrodes (e.g., 8-12) on different areas of the user's body and then measuring each electrode to obtain an electrocardiogram (ECG) signal. Conversely, wearable computing devices such as smartwatches are typically limited to contacting a single area of ​​the user's body (e.g., electrodes on the bottom of the watch face, such that the electrodes contact the user's wrist, etc.).

[0022] The inherent limitations of wearable computing devices further exacerbate this problem. Specifically, wearable computing devices are typically constructed with a minimum form factor and are therefore subject to constraints typical of such form factors (e.g., reduced cost, size, power, storage, memory, etc.). Due to these limitations and constraints, biometric signals collected using wearable computing devices are often noisy and difficult to interpret.

[0023] Similarly, biometric signals collected using specially designed medical devices typically include relatively small amounts of noise. In most cases, this noise is small enough that conventional denoising techniques can correct it. For example, traditional techniques such as filters can be used to reduce the noise typically present in ECG signals collected with specially designed medical ECG devices. However, biometric signals collected using wearable computing devices contain much higher levels of noise, and conventional denoising techniques are often insufficient.

[0024] Machine learning techniques have been successfully applied to various technologies to reduce noise that traditional techniques cannot adequately reduce. Specifically, machine learning models trained through supervised learning for noise reduction have proven particularly effective in reducing signal noise. However, because biometric sensors incorporated into wearable computing devices are a recent development, there is insufficient existing training data to train models to reduce noise present in biometric signals collected by wearable computing devices. Therefore, sufficient training data to train such models is extremely needed.

[0025] Therefore, this paper describes an implementation that proposes a machine learning model for denoising biometric signals collected using mobile biometric sensors. Specifically, the computational system can obtain multiple clean biometric signals measured using a corresponding medical device (e.g., ECG signals collected using a medical ECG device). The computational system can generate training data by applying noise to the clean biometric signals. Specifically, for each clean biometric signal, the computational system can sample noise from a randomly selected noise source. The computational system can select a random portion of the signal and can apply the sampled noise to that portion. The computational system can process the training data with a model to obtain denoised biometric signals. The computational system can then train the model based on a loss function that evaluates the differences between the multiple denoised biometric signals and the multiple clean biometric signals.

[0026] This disclosure provides various technical effects and benefits. As an example of a technical effect and benefit, the implementation described herein fully optimizes biometric signals, enabling the analysis of biometric signals to improve user health and safety. Specifically, regular analysis of some biometric signals, such as ECG and PPG signals, can provide a variety of benefits (e.g., earlier detection of health complications, optimization of sleep or exercise habits, etc.). Noisy biometric signals must be adequately denoised before being analyzed. However, conventional denoising techniques have proven insufficient when collected using mobile biometric sensors. Therefore, the implementation described herein provides a machine learning model capable of adequately denoising biometric signals collected using mobile biometric sensors, thereby enabling the analysis of noisy biometric signals and the various benefits they provide.

[0027] Exemplary embodiments of this disclosure will now be discussed in more detail with reference to the accompanying drawings.

[0028] Example devices and systems

[0029] Figure 1A , Figure 1B and Figure 1CEach of the illustrations shows a perspective view of an example non-limiting computing device (e.g., a wearable computing device) 100 according to one or more exemplary embodiments of the present disclosure. In the exemplary embodiments described herein, wearable device 100 may constitute and / or include a wearable computing device. For example, in these or other exemplary embodiments, wearable device 100 may constitute and / or include a wearable computing device, such as a wearable physiological monitoring device that can be worn by a user (also referred to herein as a “wearer”) and / or capture one or more types of physiological data of the user (e.g., heart rate (HR) data, motion data (e.g., accelerometer data), body temperature data, respiratory rate data, blood pressure data, blood oxygen level data, deoxyribonucleic acid (DNA) data, electrical skin activity (EDA) data, stress-related data, etc.).

[0030] A wearable device 100 according to an exemplary embodiment of the present disclosure may include a display 102, an attachment component 104, a fixing component 106, and a button 108 that may be located on the side of the wearable device 100. In at least one embodiment, the sides of the display 102 may be (e.g., mechanically, operatively) coupled to the attachment component 104. In some embodiments, the fixing component 106 may be located on, (e.g., mechanically, operatively) coupled to, and / or integrated with the attachment component 104. In these or other embodiments, the fixing component 106 may be positioned opposite the display 102 on the opposite end of the attachment component 104. In some embodiments, the button 108 may be located on the side of the wearable device 100, below the display 102.

[0031] The display 102 according to the example embodiments described herein may be configured and / or include any type of electronic display or screen known in the art. For example, in some embodiments, the display 102 may be configured and / or include a liquid crystal display (LCD) or an organic light-emitting diode (OLED) display, such as a transmissive LCD or a transmissive OLED display. The display 102 according to the example embodiments may be configured to provide brightness, contrast, and / or color saturation characteristics according to display settings that may be maintained by the control circuitry and / or other internal components and / or circuitry of the wearable device 100. In some embodiments, the display 102 may be configured and / or include a touchscreen, such as a capacitive touchscreen. For example, in these embodiments, the display 102 may be configured and / or include a surface capacitive touchscreen or a projected capacitive touchscreen, which may be configured to respond to contact with a charge-retaining member or tool, such as a human finger.

[0032] In some embodiments, the display 102 may be configured to provide (e.g., render) various information, such as time, date, bodily signals (e.g., physiological data of a user wearing wearable device 100), readings based on user input, and / or other information. In one embodiment, such bodily signals may include, but are not limited to, heart rate data (e.g., heart rate per minute), motion data (e.g., movement data, accelerometer data), blood pressure data, body temperature data, respiratory rate data, blood oxygen level data, deoxyribonucleic acid (DNA) data, electrical activity of the skin (EDA) data, stress-related data, and / or any other bodily signals that a person skilled in the art would understand can be measured by a wearable device such as wearable device 100. In some embodiments, readings based on user input may include, but are not limited to, the number of steps taken by the user, the distance traveled by the user, the user's sleep schedule, the user's route, the altitude climbed by the user, and / or any other metrics that a person skilled in the art would understand can be input by the user into a wearable device such as wearable device 100.

[0033] In some embodiments, the display 102 may be configured to display a low-power operation domain interface or a high-performance operation domain interface, or switch between the two. In some implementations, the low-power operation domain interface may be similar to or the same as the high-performance operation domain interface. For example, both the low-power and high-performance operation domain interfaces may display the same interface elements, but the high-performance operation domain interface may utilize higher brightness parameter values ​​(e.g., make the display brighter) and may update real-time data elements at a higher update frequency than the low-power operation domain interface.

[0034] Alternatively, in some implementations, the low-power operating domain interface may differ from the high-performance operating domain interface. For example, to minimize power consumption, the low-power operating domain interface may include interface elements that can be fully operated via the low-power processor of the wearable device 100, such as time indicators, weather indicators, date indicators, etc. Conversely, the high-performance operating domain interface may include multiple real-time data elements depicting real-time data, such as biometric data (e.g., the user's heart rate, motion information, etc.), stock quotes, calendar information (e.g., scheduled meetings, etc.), etc.

[0035] When the wearable device 100 operates in a low-power operating domain, the low-power operating domain interface can be displayed on the display 102. The low-power operating domain can be implemented when the user is not interacting with the wearable device 100, or is not anticipated to be interacting with it. Thus, the low-power operating domain interface can typically be displayed via the display 102 when the user is not directly observing the wearable device 100. In other words, the low-power operating domain interface can be displayed when the user is not actively looking at the wearable device 100.

[0036] In at least one embodiment of this disclosure, the aforementioned body signals and / or readings based on user input can be used to perform further analysis to provide the user with data such as fitness scores, sleep quality scores, calories burned by the user, and / or other data. In some embodiments, the wearable device 100 may receive (e.g., capture, collect, receive, measure) external data independently of the user, such as: the ambient temperature of the environment surrounding and / or the external environment of the wearable device 100; the amount of sunlight exposure experienced by the wearable device 100; the atmospheric pressure of the environment surrounding and / or the external environment of the wearable device 100; the air quality of the environment surrounding and / or the external environment of the wearable device 100; the location of the wearable device 100 based on, for example, a Global Positioning System (GPS); and / or other external factors that a wearable device such as the wearable device 100 may receive (e.g., capture, collect, receive, measure) as will be understood by those skilled in the art.

[0037] The attachment component 104 according to the example embodiments described herein can be used to attach (e.g., attach, fasten) the wearable device 100 to a user of the wearable device 100. In some embodiments, the attachment component 104 may take the form of, for example, a strip, elastic band, cord, and / or any other form of attachment that a person skilled in the art will understand can be used to attach a wearable device, such as wearable device 100, to a user.

[0038] The securing component 106 according to an exemplary embodiment of this disclosure can facilitate attachment of the attachment component 104 to a user of the wearable device 100. In some embodiments, the securing component 106 may include, but is not limited to, pin and hole locking mechanisms (e.g., buckles), magnetic systems, locks, clips, and / or any other type of fastener that a person skilled in the art will understand can be used to facilitate attachment of a wearable device, such as wearable device 100, to a user. In one embodiment, the wearable device 100 does not include the securing component 106. For example, in this or another embodiment, the wearable device 100 may be secured to the user with a strap that can be attached to the user's wrist and / or another suitable accessory.

[0039] Button 108, according to the example embodiments described herein, allows a user to interact with wearable device 100 and / or allows a user to provide some form of input to wearable device 100. Figure 1A , 1B In the example embodiments depicted in 1C, a button 108 is shown on the wearable device 100. However, it should be understood that the wearable device 100 is not limited thereto. For example, in some embodiments, the wearable device 100 may include any number of buttons that allow the user to further interact with the wearable device 100 and / or provide alternative inputs. In at least one embodiment, the wearable device 100 does not include the button 108. For example, as described above, in the example embodiments, the wearable device 100 may include a screen, such as a touchscreen that can receive input via (e.g., by means of) a user's touch. In additional or alternative embodiments, the wearable device 100 may include a microphone that can receive input via (e.g., by means of) a user's voice commands.

[0040] Some implementations of this disclosure are described in the context of a user's intent to perform a specific type of interaction with wearable device 100 that requires wearable device 100 to operate in a high-performance operating domain. In some implementations, this specific type of interaction may include interaction with button 108. Alternatively, in some implementations, interaction with button 108 is considered a specific type of interaction depending on which actions (if any) are performed in response to button 108 being pressed. For example, if a game application is selected and the user interacts with button 108 (and thus executes the application), then the interaction may be considered a specific type of interaction. As another example, if a "sleep" feature is highlighted and the user interacts with button 108, then the user's interaction with button 108 may not be considered a specific type of interaction.

[0041] Alternatively, in some implementations, the device can switch to a high-performance operating domain independently of user input (e.g., increasing screen brightness, modifying the device's processor operating parameters (e.g., clock speed, available cores, virtualization instances, etc.)). For example, if biometric signals (e.g., ECG signals, PPG signals, etc.) obtained via the device are denoised and analyzed to identify health status, the device can switch to a high-performance operating domain to alert the user.

[0042] In some embodiments, the wearable device 100 may constitute a portable computing device, which may be designed such that it can be inserted into a wearable housing (e.g., as shown in the image). Figure 1A , 1B(As illustrated in the example embodiments depicted in 1C). In some embodiments, the wearable device 100 may constitute a portable computing device that can be designed to be inserted into one or more of a plurality of different wearable housings (e.g., wristband housing, belt clip housing, pendant housing, housing configured to be attached to a piece of exercise equipment such as a bicycle). The wearable device 100 according to the embodiments described herein may be formed in one or more shapes and / or sizes to allow coupling to (e.g., attaching to, wearing to, carrying to) a user's body or clothing. In some embodiments, the wearable device 100 may constitute a portable computing device that can be designed to be worn in a limited manner, such as a computing device that is non-removably integrated into a wristband and / or may be intended to be worn specifically on a person's wrist (or possibly ankle).

[0043] Regardless of configuration, a wearable device 100 according to an exemplary embodiment of this disclosure may include one or more physiological and / or environmental sensors (e.g., internal physiological sensor 143, external physiological sensor 145, and / or environmental sensor 155) that can be configured to collect physiological and / or environmental data according to various embodiments disclosed herein. In some embodiments, the wearable device 100 may be configured to analyze and / or interpret the collected physiological and / or environmental data to perform one or more health, wellness, and / or well-being assessments (e.g., physical, mental, emotional, behavioral, and / or sleep quality assessments) of a user (e.g., a wearer) of the wearable device 100, according to one or more embodiments described herein. In additional and / or alternative embodiments, the wearable device 100 may be configured to communicate with another computing device or server that can perform one or more health, wellness, and / or well-being assessments (e.g., physical, mental, emotional, behavioral, and / or sleep quality assessments) of a user (e.g., a wearer) of the wearable device 100, according to one or more embodiments described herein.

[0044] Wearable device 100 according to one or more exemplary embodiments of this disclosure may include one or more physiological and / or environmental components and / or modules that may be designed to determine one or more physiological and / or environmental indicators associated with a user (e.g., a wearer) of wearable device 100. In at least one embodiment, such physiological and / or environmental components and / or modules may constitute and / or include one or more physiological and / or environmental sensors. For example, although in Figure 1A , Figure 1B and Figure 1CNot depicted in the example embodiments shown, but in some embodiments, the wearable device 100 may include one or more physiological and / or environmental sensors, such as an accelerometer, a heart rate sensor (e.g., a photoplethysmography (PPG) sensor), an electrical activity of the skin (EDA) sensor, an electrocardiogram (ECG) sensor, a body temperature sensor, an ambient temperature sensor, and / or another physiological and / or environmental sensor. In these or other embodiments, such physiological and / or environmental sensors may be disposed on, coupled to, or otherwise associated with the bottom and / or back side of the wearable device 100 (e.g., back 134).

[0045] In some embodiments, the aforementioned physiological and / or environmental sensors may be disposed on, coupled to, and / or otherwise associated with the wearable device 100, such that the sensors may come into contact with or substantially contact with human skin when the user wears the wearable device 100. For example, in embodiments where the wearable device 100 may be worn on a user's wrist, the physiological and / or environmental sensors may be disposed on, coupled to, and / or otherwise associated with the back 134, which may substantially face away from the display 102 and touch the user's arm. In one embodiment, the aforementioned physiological and / or environmental sensors may be disposed on the interior or skin side of the wearable device 100 (e.g., the side of the wearable device 100 that contacts, touches, and / or faces the user's skin, such as the back 134 and / or the bottom 142), coupled to, and / or otherwise associated with the interior or skin side. In another embodiment, physiological and / or environmental sensors may be disposed on one or more sides of the wearable device 100, including the skin side (e.g., back 134, bottom 142) and one or more sides of the wearable device 100 that face and / or are exposed to the surrounding environment (e.g., the external environment around the wearable device 100) (e.g., first side 136, second side 138, top 140, display 102).

[0046] Figure 2This is a block diagram of an environment suitable for implementing machine learning models for denoising biometric signals collected using mobile biometric sensors, according to some implementations of this disclosure. In some implementations, computing environment 200 includes computing system 202. Computing system 202 may be a system for training, providing, or otherwise implementing various machine learning models for reducing noise in biometric signals, and may include processor device 204 and memory 206. In some implementations, computing system 202 may be a computing system including multiple computing devices. Alternatively, in some implementations, computing system 202 may be one or more computing devices within computing environment 200, which includes multiple distributed devices and / or systems. Similarly, processor device 204 may include any computing or electronic device capable of executing software instructions to implement the functions described herein.

[0047] Memory 206 may be, or otherwise include, any means capable of storing data, including but not limited to volatile memory (random access memory, etc.), non-volatile memory, and storage devices (e.g., hard disk drives, solid-state drives, etc.). Specifically, memory 206 may include containerized units of software instructions (i.e., “packaged containers”). Containerized units of software instructions may collectively form a container that has been packaged using any type or manner of containerization technology.

[0048] A containerized unit of software instructions may include one or more applications and may further implement any software or hardware necessary to execute the containerized unit of software instructions in any type or manner of computing environment. For example, a containerized unit of software instructions may include software instructions (e.g., applications, dependencies, configuration files, libraries, related binaries, etc.) containing or otherwise implementing all components necessary for process isolation in any environment.

[0049] In some implementations, the memory 206 of the computing system 202 may include a machine learning module 208. The machine learning module 208 may train, instantiate, deinstantiate, optimize, store, or otherwise implement various machine learning models. For this purpose, the machine learning module 208 may include a model repository 210. The model repository 210 may store untrained, partially trained, or fully trained machine learning models. The model repository 210 may also catalog information related to the machine learning models (e.g., version information, performance metrics, predicted computational resource costs, etc.).

[0050] Model repository 210 may include machine learning models 212 for reducing noise in biometric signals. For example, model repository 210 may include a machine learning ECG denoising model 212A trained to reduce noise in ECG signals collected using mobile biometric sensors. As another example, model repository 210 may include a machine learning PPG denoising model 212B trained to reduce noise in PPG signals collected using mobile biometric sensors. Still another example, model repository 210 may include a machine learning multimodal denoising model 212C trained to reduce noise in multiple types of biometric signals collected using mobile biometric sensors.

[0051] The machine learning module 208 may include a model training module 214. The model training module 214 may be configured to train a machine learning model for use by other computing devices within the computing system 202 or computing environment 200. For example, the model training module 214 may use training data to train a model for use in wearable computing devices, mobile computing devices, etc.

[0052] Model training module 214 may include training data 216. Training data 216 may include or describe training examples for supervised learning, data for unsupervised learning, etc. Specifically, in some implementations, training data 216 may include training examples 218 for supervised training of machine learning model 212. Training examples 218 may include noisy biometric signals 220 (e.g., ECG signals, PPG signals, etc.) and corresponding clean biometric signals 222 used as ground truth signals. Model training module 214 may use training data 216 to train machine learning model 212.

[0053] Specifically, model training module 214 can train a machine learning model to transform the noisy biometric signal 220 from training examples 218 into its corresponding clean biometric signal 222 (e.g., the "true value" signal). For example, model training module 214 can process the noisy ECG signal from the noisy biometric signal 220 using a machine learning ECG denoising model 212A to generate a denoised biometric signal. Model training module 214 can train machine learning ECG denoising model 212A based on the difference between the denoised biometric signal and the clean true value biometric signal of the training examples.

[0054] To generate training data 216, model training module 214 may include training data generator 224. Training data generator 224 may include clean biometric signal information 226. Clean biometric signal information 226 may include or otherwise describe some or all of the clean biometric signals 222 of training examples 218. Each of the clean biometric signals 222 may include a relatively small amount of noise (e.g., noise below a noise threshold) or be completely noise-free. Alternatively, in some implementations, clean biometric signal information 226 may include or otherwise describe clean biometric signals captured using a medical device (e.g., a medical biometric sensor). For example, clean biometric signal information 226 may describe ECG signals captured using a medical ECG sensing device that measures ECG signals from multiple electrodes placed at different locations on a patient's body. As another example, clean biometric signal information 226 may describe PPG signals captured using a medical PPG sensing device that measures PPG signals from a sensor placed on a patient's fingertip. By using biometric signals that are known to be “clean,” such as those captured using specially designed medical devices, the implementation described in this paper can train a denoising model to denoise noisy biometric signals, so that the denoised signal can simulate a “clean” signal in terms of accuracy, noise content, etc.

[0055] Therefore, the clean biometric signal 222, described by the clean biometric signal information 226, can be used as a "true value" example in training example 218 for model 212 to simulate. Additionally, by adding noise to the clean biometric signal 222, it can be used as a source of the corresponding noisy biometric signal 220 for training example 218. More specifically, training a machine learning denoising model, such as machine learning model 212, can be done accurately and efficiently using supervised learning techniques. Supervised learning typically refers to a training process where the machine learning model generates an output, which is compared to a known correct "true value" output. Model performance can improve over time by modifying the model's parameters to reduce the difference between the model output and the true value output. However, the vast majority of biometric signals have traditionally been collected using specially customized medical devices (e.g., ECG machines, PPG machines, etc.), which typically do not capture significant amounts of noise. Furthermore, the noise present in such signals can usually be denoised using conventional techniques. Therefore, there is currently a lack of sufficient training data to train machine learning models to denoise biometric signals collected using mobile biometric sensors.

[0056] Therefore, the training data generator 224 can generate synthetic “noisy” training examples for the training examples 218 by applying sampled noise to the clean biometric signal information 226. For this purpose, the training data generator 224 may include a noise source selector 228. The noise source selector 228 may include or otherwise access multiple noise sources 230A–230N (typically noise source 230). As described herein, “noise” in a signal generally refers to unwanted, unintended, irrelevant interference or fluctuations that mask or distort the original information carried by the signal. Noise can originate from a variety of sources, including electrical interference, environmental factors, defects in equipment, and inherent limitations of the signal transmission medium.

[0057] Noise source 230 may include noise of any type or manner. In some implementations, noise source 230 may include white noise (e.g., broadband noise), pink noise (e.g., noise concentrated in the lower frequency range of the spectrum), Brownian noise (e.g., noise even more strongly concentrated in the lower frequency range of the spectrum), etc. For example, noise source 230 may be an algorithm configured to replicate white noise in forms commonly found in various biometric signals, or may include multiple algorithms, each specifically configured to replicate white noise found in a particular type of biometric signal (e.g., an algorithm to replicate white noise found in ECG readings, another algorithm to replicate white noise found in PPG readings, etc.).

[0058] Alternatively or concurrently, in some implementations, noise source 230 may include a signal conditioning process. The signal conditioning process may refer to the process of modifying the characteristics of the biometric signal to simulate noise that might be captured by a mobile biometric sensor. For example, noise source 230 may include a signal compression process that compresses a portion of the clean biometric signal. As another example, noise source 230 may include a signal expansion process that extends a portion of the clean biometric signal. Yet another example, noise source 230 may include a signal trimming process that prunes (i.e., removes) a portion of the clean biometric signal.

[0059] Alternatively or concurrently, in some implementations, noise source 230 may include noise sampled from noisy biometric signals collected using a medical device. Specifically, in some implementations, in addition to clean biometric signal information 222, training data generator 224 may obtain noisy biometric signal information captured using the same (or similar) medical device used to capture clean biometric signal information 222. For example, noisy biometric signal information can be captured by identifying a relatively small amount of existing noise (if present) present in the extracted clean biometric signal information 222. As another example, noisy biometric signal information may be a noisy signal caused by improper use of a medical device, use of an older or obsolete medical device, etc.

[0060] Noise source selector 228 can select one (or more) of noise sources 230 to be applied to some (or all) of the clean biometric signal 222 described by clean biometric signal information 226. For this purpose, noise source selector 228 can perform a noise selection process. In some implementations, the noise selection process can be partially or completely random. For example, the noise selection process can be a randomized noise selection process that first identifies a random number of noise sources to be selected, and then randomly selects that number of noise sources from the noise sources 230. Alternatively, in some implementations, the noise selection process can be a non-random or deterministic noise selection process. For example, the noise selection process can evaluate the clean biometric signal 222 described by clean biometric signal information 226, and then select a specific noise source from the noise sources 230 based on that evaluation.

[0061] In some implementations, the noise source selector 228 may select the number and / or type of noise source 230 for each of the clean biometric signals 222 described by the clean biometric signal information 226 on a signal-by-signal basis (e.g., repeating the noise selection process once for each of the noisy biometric signals 220). Alternatively, in some implementations, the noise source selector 228 may select the number and / or type of noise source 230 for multiple portions of each of the clean biometric signals 222 (e.g., repeating the noise selection process multiple times to apply noise to multiple portions of each of the noisy biometric signals 220). Alternatively, in some implementations, the noise source selector 228 may select the number and / or type of noise source 230 for certain sets of clean biometric signals 222, or may select a set of noise sources to be applied to all the clean biometric signals 222 described by the clean biometric signal information 226 (e.g., repeating the noise selection process once or multiple times for one or more sets of noisy biometric signals 220).

[0062] The training data generator 224 may include a noise sampler 232. The noise sampler 232 may sample noise from a noise source 230 selected by the noise source selector 228. Specifically, the noise sampler 232 may generate sampled noise information 234 describing the noise sampled from the noise source 230. The noise sampler 232 may use any type or manner of conventional noise sampling techniques to sample noise from the noise source 230.

[0063] Training data generator 224 may include a signal portion selector 236. Signal portion selector 236 may select one of the clean biometric signals 222. Signal portion selector 236 may identify the portion (or all) of the signal for which sampled noise, as described by sampled noise information 234, is applied. In some implementations, signal portion selector 236 may select multiple portions of a signal from the clean biometric signal 222 to apply sampled noise (e.g., a set of the same sampled noise, a set of different sampled noise, etc.). Alternatively, in some implementations, signal portion selector 236 may select the same portion or portions of the same type of signal from multiple signals in the clean biometric signal 222. For example, if the clean biometric signal 222 comprises an ECG signal of at least five seconds in length, signal portion selector 222 may select the first two seconds of multiple signals from the clean biometric signal 222.

[0064] The training data generator 224 may include a noise applicator 238. The noise applicator 238 may apply sampled noise to the identified portion of the clean biometric signal 222 to generate a corresponding noisy biometric signal 220. In this way, the training data generator 224 can sample and apply noise to the clean biometric signal 222 to obtain the corresponding noisy biometric signal 220, thereby creating training examples 218 described by the training data 216 for supervised learning of the machine learning model 212.

[0065] In some implementations, training data generator 224 can generate a single training example from each clean biometric signal 222 described by clean biometric signal information 226. In other words, training data generator 224 can create a total number of training examples equal to or less than the number of clean biometric signals 222 described by clean biometric signal information 226. For example, suppose the clean biometric signal 222 includes three biometric signals. Noise source selector 228 can select some (or all) of noise sources 230, and noise sampler 232 can sample noise from the selected noise sources 230 to generate sampled noise information 234. Signal portion selector 236 can select a portion of the first clean biometric signal, and noise applicator 238 can apply the sampled noise to the selected portion of the clean biometric signal to generate a corresponding first noisy biometric signal 220. This process can be repeated for the second and third clean biometric signals to generate corresponding second and third noisy biometric signals.

[0066] Alternatively, in some implementations, the training data generator 224 can generate multiple training examples from a single clean biometric signal described by the clean biometric signal information 226. Following the previous example, instead of immediately repeating the noise application process for the second and third clean biometric signals, the training data generator can generate another noisy biometric signal from the first clean biometric signal by applying different noise to the first clean biometric signal and / or selecting different portions of the clean biometric signal for noise application. For example, after generating a first noisy biometric signal by applying sampled noise to the first clean biometric signal, the training data generator can generate a second noisy biometric signal from the first clean biometric signal by selecting different noise (e.g., selecting different sources, sampling the sources in different ways, etc.) and / or by selecting different portions of the first clean biometric signal and subsequently applying sampled noise to the selected portions of the first clean biometric signal to generate a second noisy biometric signal. In this way, by generating multiple training examples from the same clean biometric signals, the implementation described in this paper can generate a rich training dataset from a small number of ground truth training examples.

[0067] The model training module 214 may include a model optimizer 240. The model optimizer 240 can perform various model training and / or optimization operations, such as evaluating a loss function, adjusting the parameters or weights of model 212, quantizing model weights, performing distillation training, etc. Specifically, the model optimizer 240 may include a loss function 242. The loss function 242 may include a loss function or "optimization" function configured to evaluate the model output for the purpose of training the model.

[0068] In some implementations, the loss function may include multiple loss functions for evaluating the output of a model trained to denoise multiple corresponding types of biometric signals. For example, one of the loss functions 242 may be configured to train a model to denoise ECG signals, while another of the loss functions 242 may be configured to train a model to denoise PPG signals. The loss function 242 may be evaluated by a model optimizer 240, and based on this evaluation, the model optimizer 240 may adjust the parameters of the model evaluated by the loss function 242. The model optimizer 240 and the loss function 242 will be related to... Figure 3 Let's discuss this in more detail.

[0069] In some implementations, model optimizer 240 can perform various optimization tasks for the trained model (e.g., distillation training, quantization, fine-tuning, storage, indexing, transferring the model to other devices, versioning implementation, etc.). In some implementations, model optimizer 240 can perform quantization processes on various models of machine learning model 212 after training the model. For example, model optimizer 240 can reduce the precision of model parameters (e.g., from single-precision floating-point (FP32) to half-precision floating-point (FP16.16) or 8-bit integer (INT8), etc.), and thus reduce the computational resources required to utilize the model (e.g., memory resources, computation loops, etc.).

[0070] In another example, model optimizer 240 can perform distillation training to generate a machine learning denoising distillation model. As described herein, a “machine learning denoising distillation model” can refer to a model that has been trained on a “teacher” model to denoise biometric signals. Specifically, the distillation model can be trained on the inputs and outputs of a “teacher” model (such as machine learning model 212). The distillation model can exhibit similar (or identical) denoising performance to machine learning model 212 while utilizing fewer computational resources, making it ideal for use by mobile computing devices with limited and weaker computational resources.

[0071] In some implementations, the computing environment 200 may include a wearable computing device 244. The wearable computing device 244 may be any type or manner of wearable device that can implement biometric sensors directly (e.g., where biometric sensors are physically built into the wearable computing device 244) or indirectly (e.g., via biometric sensors communicatively coupled to the wearable computing device 244). For example, the wearable computing device may be... Figure 1A-1C Wearable device 100. Other examples of wearable computing device 244 may include smartwatches, biometric tracking devices worn on the wrist or other appendages, biometric sensors, distributed collections of sensors and devices, augmented reality (AR) or virtual reality (VR) devices, glasses, clothing devices, etc.

[0072] The wearable computing device 244 may include a processor device 246 and a memory 248. The processor device 246 and the memory 248 will... Figure 4This will be discussed in more detail. Wearable computing device 244 may include mobile biometric sensor 250. As described herein, "mobile" biometric sensor generally refers to any type or manner of biometric sensor that is not limited to a specific, specially tailored medical device or medical context. For example, conventional ECG machines typically require multiple electrodes to be attached to a patient to capture biometric signals. Furthermore, conventional ECG machines are relatively stationary, and their use is generally limited to healthcare providers. In contrast, a mobile ECG sensor incorporated into wearable computing device 244 may incorporate a single electrode that contacts the patient's skin when worn, thereby enabling the sensor to periodically capture biometric signals from the user.

[0073] Specifically, the wearable computing device 244 may integrate with or otherwise communicatively access the mobile biometric sensor 250. In some implementations, the mobile biometric sensor 250 may be integrated into the wearable computing device 244 to contact the skin of the user wearing the wearable computing device 244. For example, if the wearable computing device 244 is a smartwatch device, the mobile biometric sensor 250 may be located opposite the face of the watch, i.e., on the lower or “bottom” surface of the watch that contacts the user’s wrist. In this way, the mobile biometric sensor 250 can measure various biometric signals from the user via the user’s surface in contact with the mobile biometric sensor 250.

[0074] The memory 248 of the wearable computing device 244 may include a biometric analysis module 252. In some implementations, the biometric analysis module 252 may include a machine learning biometric denoising model 254. The machine learning biometric denoising model 254 may be a model trained to reduce noise present in biometric signals measured using the mobile biometric sensor 250. For example, the machine learning biometric denoising model 254 may be one or more of machine learning models 212 trained at the computing system 202. As another example, the machine learning biometric denoising model 254 may be one or more models created based on the machine learning model 212 (e.g., trained via distillation, quantization, etc.). In some implementations, the memory 248 may include a model training module 255 that can train the machine learning biometric denoising model 254 as described with respect to the model training module 214 of the computing system 202.

[0075] Biometric analysis module 252 can acquire noisy biometric signal 256 via mobile biometric sensor 250. In some implementations, biometric analysis module 252 can process noisy biometric signal 256 using machine learning biometric denoising model 254 to obtain denoised biometric signal 258. Denoised biometric signal 258 may include less noise than noisy biometric signal 256 while retaining most (or all) of the signal present in noisy biometric signal 256. For example, if noisy biometric signal 256 is an ECG signal, then denoised biometric signal 258 may be an ECG signal exhibiting reduced noise while retaining consistency between ECG signal characteristics (e.g., P-wave, etc.). Alternatively, in some implementations, biometric analysis module 252 may provide noisy biometric signal 256 to computing system 202 for denoising (e.g., using machine learning model 212, etc.).

[0076] The biometric analysis module 252 may include a signal analyzer 260. In some implementations, the signal analyzer 260 may analyze the denoised biometric signal 258 to obtain biometric measurements 262 (e.g., physiological values ​​or measurements). Alternatively, in some implementations, the signal analyzer 260 may provide the denoised biometric signal 258 to the computing system 202 for analysis. In some implementations, the signal analyzer 260 may cause information to be displayed or otherwise notified to the user of the wearable computing device 244. Various interactions implemented by the signal analyzer 260 will relate to... Figure 4 Let's discuss this in more detail.

[0077] In some implementations, computing environment 200 may include user computing device 264. User computing device 264 may include processor device 266 and memory 268 as described with respect to computing system 202 and / or wearable computing device 244. User computing device 264 may be any type or manner of computing device utilized by a user of the wearable computing device, such as smartphones, laptops, desktop computers, tablets, AR / VR devices, cloud devices, virtualization devices, automotive computing devices, etc.

[0078] The memory 268 of the user computing device 264 may include a biometric analysis module 270. The biometric analysis module may perform some or all of the functionalities described regarding the biometric analysis module 252. For example, after acquiring a noisy biometric signal 256, the wearable computing device 244 may partially or completely denoise the noisy biometric signal 256 using a machine learning biometric denoising model 254, and then provide the partially or completely denoised biometric signal to the biometric analysis module 270 of the user computing device 264. The biometric analysis module 270 may acquire a denoised biometric signal 258 (e.g., by fully denoising the partially denoised signal if necessary). The biometric analysis module 270 may analyze the denoised biometric signal 258 to obtain a biometric measurement 262, and may provide the biometric measurement 262 to the wearable computing device 244.

[0079] Figure 3 This is based on some implementation methods of this disclosure. Figure 2 A block diagram of the training data generator 224 of the model training module 214 is provided. Specifically, the training data generator 224 may include a noise source selector 228. The noise source selector 228 may include noise sources 230, a random selection function 302, and a deterministic selection function 304. The random selection function 302 may be configured to select a random number of sources from the noise sources 230. The deterministic selection function 304 may select specific sources from the noise sources 230 based on source selection information 306. The source selection information 306 may map certain noise sources 230 to the corresponding noise exhibited by noisy signals collected using mobile biometric sensors. In other words, given a specific mobile biometric sensor (e.g., an ECG sensor), the source selection information 306 may identify certain sources of noise sources 230 that best match the noise found in the signal collected using that specific mobile sensor.

[0080] The training data generator 224 can use the noise source selector 228 to generate selected noise source information 308. The selected noise source information 308 can indicate the noise source selected by the noise source selector 224. The noise sampler 310 can generate sampled noise 312 by sampling noise from the noise source 230 indicated by the selected noise source information 308. The training data generator can process the clean biometric signal 222 with the signal portion selector 236 to obtain selected signal portion information 314. The selected signal portion information 314 can indicate the portion of the clean biometric signal 222 used to apply the sampled noise 312. The noise applicator 238 can apply the sampled noise 312 to the portion of the clean biometric signal 222 indicated by the selected signal portion information 314.

[0081] Figure 4 Block diagrams of the above-described example non-limiting wearable devices 100 and / or 244 according to one or more exemplary embodiments of the present disclosure are shown. That is, for example, Figure 4 Block diagrams are shown of one or more internal and / or external components of the above-described example non-limiting wearable devices 100 and / or 244 according to one or more example embodiments of the present disclosure.

[0082] As referenced above Figure 1A-1C , Figure 2 and Figure 3 As illustrated in the example embodiments, the wearable device 100 may constitute and / or include a wearable computing device, such as a wearable physiological monitoring device. For example, in Figure 4 In the example embodiments depicted, wearable device 100 may constitute and / or include a wearable physiological monitoring device that may be worn by user 10 (also referred to herein as "wearer" or "wearer 10") and / or may be configured to collect data about activities performed by user 10 and / or about user 10's physiological state (e.g., temperature, blood volume, pulse, pulse oxygenation, heart rate, heart rhythm, etc.). In this embodiment or another embodiment, such data may include data representing the surrounding environment of user 10 or user 10's interaction with the environment. For example, in some embodiments, the data may constitute and / or include motion data about user 10's movement, ambient light, ambient noise, air quality, and / or physiological data obtained by measuring various physiological characteristics of user 10 (e.g., heart rate, respiratory data, body temperature, blood oxygenation level, sweating level, movement data).

[0083] While certain embodiments are disclosed herein in the context of wearable physiological monitoring devices, it should be understood that this disclosure is not limited thereto. For example, it should be understood that one or more embodiments may be performed and / or implemented using any suitable or desired type of computing device or combination of computing devices, such as client computing devices, laptop computers, tablet computers, wearable computing devices (e.g., wearable device 100), smartphones, and / or other computing devices, whether or not the computing devices are wearable.

[0084] like Figure 4As shown, a wearable device 100 according to an example embodiment of this disclosure may include one or more audio and / or visual feedback components 130, such as an electronic touchscreen display unit, a light-emitting diode (LED) display unit, an audio speaker, an LED lamp, a buzzer, and / or another type of audio and / or visual feedback module. In some embodiments, one or more audio and / or visual feedback modules 130 may be located on the front side of the wearable device 100 and / or on the display 102 and / or otherwise associated with the front side of the wearable device and / or the display. For example, in a wearable embodiment of the wearable device 100, an electronic display, such as the display 102, may be configured to be externally presented to a user 10 viewing the wearable device 100.

[0085] A wearable device 100 according to an exemplary embodiment of this disclosure may include a control circuitry system 110. Although in Figure 4 The illustrations show certain modules and / or components as part of the control circuitry system 110, but it should be understood that the control circuitry system 110 associated with the wearable device 100 and / or other components or devices according to exemplary embodiments of this disclosure may include additional components and / or circuitry systems, such as... Figure 4 One or more additional components are shown among the components depicted. Furthermore, in some embodiments, one or more of the shown components of the control circuitry system 110 may be omitted and / or differ from those shown. Figure 4 The components shown and described in association with it.

[0086] The term "control circuit system" is used herein in its broad and / or general sense, and may include any combination of software and / or hardware elements, devices, and / or features that can be implemented in relation to the operation of wearable device 100. Furthermore, in certain contexts herein, the term "control circuit system" may be used substantially interchangeably with one or more of the terms "controller," "integrated circuit," "IC," "application-specific integrated circuit," "ASIC," "controller chip," etc.

[0087] The control circuitry system 110 according to an exemplary embodiment of this disclosure may be configured and / or include one or more processors, data storage devices, and / or electrical connections. In one embodiment, the control circuitry system 110 may be implemented on a system-on-a-chip (SoC); however, those skilled in the art will recognize that other hardware and / or firmware implementations are possible.

[0088] In one or more embodiments of this disclosure, the control circuitry system 110 may comprise and / or include one or more processors 181 configured to execute computer-readable instructions that, when executed, cause the wearable device 100 to perform one or more operations. In at least one embodiment, the control circuitry system 110 may comprise and / or include a processor 181 configured to execute operational code (e.g., instructions, processing threads, software) of the wearable device 100, such as firmware. The processors 181 according to the example embodiments described herein may each be processing means. For example, in Figure 4 In the example embodiments depicted herein, processor 181 may each be a central processing unit (CPU), a microprocessor, a microcontroller, an integrated circuit (e.g., an application-specific integrated circuit (ASIC)), and / or another type of processing device. In this example embodiment or another example embodiment, processor 181 may be coupled (e.g., electrically, communicatively, physically, operatively) to one or more components of control circuitry system 110 and / or wearable device 100, such that processor 181 may facilitate one or more operations according to one or more example embodiments described herein.

[0089] In at least one embodiment of this disclosure, the aforementioned computer-readable instructions and / or operation codes, executable by processor 181, may be stored in one or more data storage devices of the wearable device 100. Figure 4 In the example embodiments depicted herein, such computer-readable instructions and / or operational codes may be stored in the memory 183 of the wearable device 100. In this example embodiment or another example embodiment, the memory 183 may be coupled (e.g., electrically, communicatively, physically, operatively) to the control circuitry system 110 and / or one or more components of the wearable device 100, such that the memory 183 may facilitate one or more operations according to one or more example embodiments described herein.

[0090] Memory 183 according to the example embodiments described herein may store computer-readable and / or computer-executable entities (e.g., data, information, applications, models, algorithms) that can be created, modified, accessed, read, retrieved, and / or executed by each processor in processor 181. In some embodiments, memory 183 may be configured, included, (e.g., operatively) coupled to a computing system and / or medium, such as one or more computer-readable media, volatile memory, non-volatile memory, random access memory (RAM), read-only memory (ROM), hard disk drive, flash drive, and / or other memory devices, and / or otherwise associated with said computing system and / or medium. In these or other embodiments, such one or more computer-readable media may include, be configured, (e.g., operatively) coupled to one or more non-transitory computer-readable media, and / or otherwise associated with one or more non-transitory computer-readable media. Although not explicitly stated in the text... Figure 4 The example embodiments shown are illustrated, but in some embodiments, memory 183 may include (e.g., storing) biometric analysis module 111, and / or other modules and / or data that may be used to facilitate one or more operations described herein.

[0091] The control circuit system 110 according to an exemplary embodiment of the present disclosure may be configured and / or include a biometric analysis module 111. The biometric analysis module 111 according to an exemplary embodiment of the present disclosure may be configured and / or include one or more hardware and / or software components and / or features that may be configured to perform a switch from one operating domain to another, or a transition from one operating domain to another, according to one or more embodiments described herein. For example, in some embodiments, the biometric analysis module 111 may be configured and / or include... Figure 2 Biometric analysis module 252.

[0092] In some embodiments, the biometric analysis module 111 may receive noisy biometric signals (e.g., via internal physiological sensors 143, external physiological sensors 145, environmental sensors 155, etc.), denoise the received signals, and generate biometric measurements based on the denoised signals. At least in part based on (e.g., in response to) obtaining a specific biometric measurement, the wearable device 100 may perform one or more operations described herein to promote changes (e.g., improvements) in the user 10's health, well-being, and / or well-being (e.g., physical, psychological, emotional, behavioral, and / or sleep quality).

[0093] For example, in at least one embodiment, the wearable device 100 may perform operations that may include, but are not limited to: presenting biometric measurements to user 10 and / or another computing device; providing an interpretation of the biometric measurements to user 10 and / or another computing device, the interpretation of which may include activities as defined herein, suggesting one or more health improvement recommendations and / or causing another computing device to make such recommendations at least in part based on (e.g., using) the biometric measurements (e.g., recommendations for user 10 to seek medical care or seek epidemiological testing); implementing one or more health-promoting features at least in part based on (e.g., using) relevance or irrelevance and / or causing another computing device to implement such features (e.g., vibrating in a particular manner to indicate to the user that they should reduce physical exertion, etc.); and / or another operation according to one or more example embodiments of this disclosure.

[0094] In some embodiments, the biometric analysis module 111 may be communicatively coupled to one or more internal physiological sensors 143 that may be embedded in and / or integrated into the wearable device 100. In some embodiments, the biometric analysis module 111 may optionally communicate with one or more external physiological sensors 145 that are not embedded in and / or integrated into the wearable device 100 (e.g., electrodes or sensors integrated into another electronic device). In some embodiments, examples of the internal physiological sensors 143 and / or external physiological sensors 145 may constitute and / or include, but are not limited to, one or more sensors capable of measuring (e.g., capturing, collecting, receiving) physiological data of the user 10, such as heart rate, blood oxygen level, movement, respiration, sweating, stress data, and / or other physiological data of the user 10.

[0095] exist Figure 4 In the example embodiments depicted, the wearable device 100 may include one or more data storage components 151 (in... Figure 4(Represented as "Data Storage 151" in the text). Data storage component 151 according to the example embodiment can be configured and / or include any suitable or desired type of data storage, such as solid-state memory that may be volatile or non-volatile. In some embodiments, such solid-state memory of the wearable device 100 can be configured and / or include any of a variety of technologies, such as flash memory integrated circuits, phase-change (PC) memory, phase-change (PC) random access memory (RAM), programmable metallized cell RAM (PMC-RAM or PMCm), bidirectional unified memory (OUM), resistive RAM (RRAM), NAND memory, NOR memory, EEPROM, ferroelectric memory (FeRAM), MRAM, or other discrete NVM (non-volatile solid-state memory) chips. In some embodiments, data storage component 151 can be used to store system data, such as operating system data and / or system configuration or parameters. In some embodiments, wearable device 100 can include data storage used as a buffer and / or cache memory for operation by control circuitry system 110.

[0096] Data storage component 151 according to an example embodiment may include various sub-modules that can be implemented to facilitate physiological monitoring and the health, wellness and / or well-being assessment principles and features disclosed herein (e.g., temperature sensing) according to one or more embodiments. For example, in at least one embodiment, data storage 151 may include one or more sub-modules, which may include, but are not limited to: an information collection module (e.g., a biometric analysis module 111) that manages the collection of physiological and / or environmental data (e.g., body temperature sensing) related to any health, wellness, and / or well-being assessment described herein; a heart rate determination module that determines the value and / or pattern of one or more types of heart rate of user 10; a condition determination module that determines the condition that may cause the temperature of a surface of user 10 (e.g., user's skin), such as illness, hyperthermia, hypothermia, exercise, etc.; a presentation module that manages the presentation of information (e.g., body temperature, heart rate, heart rhythm, etc.) that may be associated with any health, wellness, and / or well-being assessment described herein to user 10; a feedback management module that collects and interprets any input data and / or feedback received from user 10; and / or another sub-module.

[0097] The wearable device 100 according to an example embodiment may further include a power storage module 153 (referred to as "power storage 153"), which may constitute and / or include a rechargeable battery, one or more capacitors, or other charge retention devices. In some embodiments, the control circuitry 110 may utilize the power stored by the power storage module 153 for operation of the wearable device 100, such as for powering the display 102. In some embodiments, the power storage module 153 may be accessible via the host interface of the wearable device 100 (e.g., via one or more host interface circuitry systems and / or component 176). Figure 4 It is referred to as “Host Interface 176” and / or receives power through other means.

[0098] The wearable device 100 according to an example embodiment may further include one or more environmental sensors 155. In at least one embodiment, examples of such environmental sensors 155 may include, but are not limited to, sensors that can determine and / or measure, for example, ambient light, external (non-body) temperature, altitude, device location (e.g., Global Positioning System (GPS)) and / or other environmental data.

[0099] The wearable device 100 according to an example embodiment may further include one or more connectivity components 170, which may include, for example, a wireless transceiver 172. The wireless transceiver 172 according to an example embodiment may be communicatively coupled to one or more antenna devices 195, which may be configured to wirelessly transmit data and / or power signals to and / or receive data and / or power signals from the wearable device 100 using, but not limited to, point-to-point, WLAN, and / or cellular communications. For example, the wireless transceiver 172 may be used to connect the wearable device 100 to an external computing device (…). Figure 4 (not shown) – such as external client computing devices (e.g., smartphones, tablets, computers) and / or external host systems (e.g., servers) that can be configured to interface with wearable device 100 – transmit data and / or power between them. In some embodiments, wearable device 100 may include one or more host interface circuit systems and / or components 176 (in... Figure 4 (referred to as "host interface 176"), such as a wired interface component, which can communicatively couple the wearable device 100 to the aforementioned external computing device (e.g., smartphone, tablet computer, computer, server) to receive data and / or power from and / or transmit data to it.

[0100] According to the example embodiment, the connectivity component 170 may further include one or more user interface components 174 (in... Figure 4The wearable device 100 may use one or more user interface components (referred to as "user interface 174") to receive input data from and / or provide output data to the user 10. In some embodiments, user interface component 174 may be (e.g., operatively, communicatively) coupled to and / or otherwise associated with audio and / or visual feedback component 130. For example, in these embodiments, the display 102 of the wearable device 100 may constitute and / or include a touchscreen display that may be configured to provide (e.g., render) output data to the user 10 and / or receive user input through contact between the user and the touchscreen display using audio and / or visual feedback component 130. In some embodiments, user interface component 174 may further constitute and / or include one or more buttons or other input components or features.

[0101] The connectivity component 170 according to the example embodiment may further include a host interface circuitry and / or component 176, which may be, for example, an interface that the wearable device 100 can use to communicate with the aforementioned external computing device (e.g., a smartphone, tablet, computer, server) via a wired or wireless connection. The host interface circuitry and / or component 176 according to the example embodiment may utilize any suitable or desired communication protocol and / or physical connector and / or otherwise associate with said communication protocol and / or physical connector, such as Universal Serial Bus (USB), Micro USB, Wi-Fi, Bluetooth, FireWire, PCIe, or the like. For wireless connectivity, the host interface circuitry and / or component 176 according to the example embodiment may be integrated with the wireless transceiver 172.

[0102] While specific functional modules and components are shown and described herein, it should be understood that a number of different approaches can be used to implement the authentication management functionality according to this disclosure. For example, in some embodiments, the control circuitry system 110 may be configured and / or include one or more processors (e.g., processor 181) that can be controlled by computer-executable instructions stored in memory (e.g., memory 183, data storage component 151) to provide functionality such as that described herein. In other embodiments, such functionality may be provided in the form of one or more specially designed electrical circuits. In some embodiments, such functionality may be provided by one or more processors (e.g., processor 181) that can be controlled by computer-executable instructions that can be stored in memory (e.g., memory 183, data storage component 151) that can be (e.g., communicatively, operably, electrically) coupled to one or more specially designed electrical circuits. Various examples of hardware that can be used to implement the concepts outlined herein may include, but are not limited to, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), and general-purpose microprocessors that can be coupled to memory storing executable instructions for controlling a general-purpose microprocessor.

[0103] Figure 5 An illustration is shown of an example non-limiting user evaluation management system 500 according to one or more example embodiments of the present disclosure. Figure 5 The user evaluation management system 500 depicted illustrates an example, non-limiting networking relationship between a wearable device 100, an external computing device 504, and / or one or more intelligent systems 512 according to one or more embodiments. Specifically, it should be noted that some or all of the user evaluation management system 500 and its components or subsystems are shown as examples of providing operations requiring a high-performance operating domain.

[0104] Refer to the above and Figure 4 The exemplary embodiments depicted herein allow the wearable device 100, according to exemplary embodiments of this disclosure, to perform one or more health, wellness, and / or well-being assessments of the user 10 (e.g., physical, mental, emotional, behavioral, and / or sleep quality assessments) and / or perform operations based on such assessments to promote changes (e.g., improvements) in the user 10's health, wellness, and / or well-being. Therefore, in some embodiments described in this disclosure, the wearable device 100 may be able to and / or be configured to collect physiological sensor readings of the user 10 and / or use such readings to perform such assessments and / or operations.

[0105] However, in additional and / or alternative embodiments, wearable device 100 and / or another electronic and / or computing device that can be used to detect physiological information of user 10 may communicate with external computing device 504. In these and / or other embodiments, external computing device 504 may be configured to use such physiological information of user 10 to perform one or more health, wellness, and / or well-being assessments of user 10 (e.g., body temperature sensing) according to one or more embodiments described herein. In these and / or other embodiments, at least in part based on (e.g., in response to) performing such assessments, external computing device 504 may perform one or more operations described herein to promote changes (e.g., improvements) in user 10's health, wellness, and / or well-being (e.g., physical, mental, emotional, behavioral, and / or sleep quality).

[0106] As described in this disclosure, a wearable device 100 according to an example embodiment can be configured to collect one or more types of physiological and / or environmental data using embedded sensors and / or external devices and to transmit or relay such information to other devices via one or more networks 506. In some embodiments, this includes relaying information to a device capable of serving as an internet-accessible data source, thereby allowing the collected data to be viewed, for example, using a web browser or a web-based application, such as at an external computing device 504. For example, when a user 10 wears the wearable device 100, the wearable device 100 can use one or more environmental and / or physiological sensors to capture, process, and / or store environmental data and / or the user 10's physiological data (e.g., heart rate, motion data, temperature, respiration, sweating, EDA, stress data). The wearable device 100 according to the example embodiment can then transmit data representing such environmental data and / or the user 10's physiological data via network 506 to an account on a web service, computer, mobile phone, and / or health station, where the data can be stored, processed, and visualized by the user 10 and / or another entity (e.g., a healthcare professional).

[0107] While the wearable device 100 is shown to have a display in the exemplary embodiments of this disclosure, it should be understood that in some embodiments, the wearable device 100 may not have any type of display unit. In some embodiments, the wearable device 100 may have audio and / or visual feedback components, such as light-emitting diodes (LEDs), buzzers, speakers, and / or displays with limited functionality. The wearable device 100 according to the exemplary embodiments may be configured to attach to the body or clothing of the user 10. For example, in these or other embodiments, the wearable device 100 may be configured as a bracelet, watch, ring, electrode, finger clip, toe clip, chest strap, ankle strap, and / or a device placed in a pocket. In additional or alternative embodiments, the wearable device 100 may be embedded in something that comes into contact with the user 10, such as clothing, a cushion, blanket, pillow, and / or another accessory that may be positioned under the user 10.

[0108] In one or more embodiments of this disclosure, communication between the wearable device 100 and the external computing device 504 may be facilitated by a network 506. In some embodiments, the network 506 may constitute and / or include one or more of, for example, an ad hoc network, a peer-to-peer communication link, an intranet, an extranet, a virtual private network (VPN), a local area network (LAN), a wireless LAN (WLAN), a wide area network (WAN), a wireless WAN (WWAN), a metropolitan area network (MAN), a portion of the Internet, a portion of the Public Switched Telephone Network (PSTN), a cellular telephone network, and / or any other type of network. In some embodiments, communication between the wearable device 100 and the external computing device 504 may also be performed via a direct wired connection. In these or other embodiments, the direct wired connection may be associated with any suitable or desired communication protocol and / or physical connector, such as Universal Serial Bus (USB), Micro USB, Wi-Fi, Bluetooth, FireWire, PCIe, etc.

[0109] In exemplary embodiments of this disclosure, various computing devices may communicate with wearable device 100 to facilitate assessments and / or changes (e.g., improvements) in the health, wellness, and / or well-being of user 10. Although in Figure 5 In the example embodiment shown, the external computing device 504 is depicted as a smartphone, but it should be understood that this disclosure is not limited thereto. For example, the external computing device 504 according to the example embodiment may be configured and / or include, for example, Figure 5 The depiction includes a smartphone, personal digital assistant (PDA), mobile phone, tablet computer, personal computer, laptop computer, smart TV, video game console, server, and / or another computing device that may be located outside the wearable device 100, all with a display 508.

[0110] Figure 5 The networking relationships depicted in the example embodiments illustrate how, in some embodiments, an external computing device 504 can be implemented to perform one or more health, wellness, and / or well-being assessments of user 10 (e.g., body temperature assessment) and / or to perform operations based on such assessments to facilitate changes (e.g., improvements) in user 10's health, wellness, and / or well-being. For example, in one embodiment, user 10 may wear a wearable device 100, which may be equipped as a bracelet with one or more physiological sensors but no display. In this and / or another embodiment, when user 10 wears wearable device 100, wearable device 100 may use physiological sensors to capture, process, and / or store environmental data and / or user 10's physiological data (e.g., temperature). Wearable device 100 according to the example embodiment can then transmit data representing such environmental data and / or user 10's physiological data via network 506 to an account on a web service, computer, mobile phone, and / or health station, where the data may be stored, processed, and visualized by user 10 and / or another entity (e.g., a healthcare professional). In some embodiments, the wearable device 100 may periodically or continuously transmit such information to an external computing device 504 via a network 506.

[0111] In additional and / or alternative embodiments, wearable device 100 may store the physiological and / or environmental data collected described above and transmit such data to external computing device 504 in response to a triggering event, such as detecting an abnormal surface temperature of user 10's skin (e.g., above or below a threshold range of standard human body temperature) after a defined activity has been performed for a period of time. In some embodiments, wearable device 100 may transmit such data to external computing device 504 in response to detecting that external computing device 504 has executed a command—e.g., manually or automatically executing instructions to synchronize the collected physiological and / or environmental data and perform one or more health, wellness, and / or well-being assessments (e.g., temperature assessments) of user 10 as described herein.

[0112] In some embodiments, the external computing device 504 may present (e.g., provide, render) possible causes of the user 10's body temperature. For example, in these or other embodiments, the external computing device 504 may generate a smart notification 510 that may include such body temperature and / or one or more health improvement suggestions (e.g., suggestions to reduce physical activity, suggestions to see a healthcare provider, etc.) that, if performed by the user 10 and / or when performed by the user, may promote changes (e.g., improvements) in the user 10's health, well-being, and / or well-being (e.g., body temperature). Figure 5In the example embodiment depicted, the external computing device 504 can render a smart notification 510 with such body temperature and health improvement suggestions on the display 508, so that the user 10 and / or another entity (e.g., a healthcare professional, mental health care professional, sleep therapy provider, doctor, caregiver) can view such information.

[0113] Although not in Figure 5 As illustrated in the example embodiments, but in some embodiments, the wearable device 100 may: sense the body temperature of the user 10; determine one or more health improvement suggestions based on (e.g., in response to) the sensed body temperature; generate a smart notification 510 that includes the body temperature and health improvement suggestions; and render the information on the display 102 of the wearable device 100.

[0114] In one embodiment of this disclosure, the wearable device 100 and / or the external computing device 504 may implement (e.g., initiate, run, operate) one or more health enhancement features that may be included in the wearable device 100 and / or the external computing device 504, such as a health enhancement audio feature (e.g., by playing a sound to alert the user to abnormal body temperature) and / or another health enhancement feature of the wearable device 100 and / or the external computing device 504.

[0115] In another embodiment of this disclosure, the wearable device 100 and / or the external computing device 504 may facilitate the implementation of one or more health enhancement features of another computing device, such as one or more intelligent systems 512. In this or another embodiment, the intelligent system 512 may constitute and / or include, but is not limited to, an audio system (e.g., a residential audio system), a lighting system (e.g., a residential lighting system), an HVAC system (e.g., a residential HVAC system), an exercise system (e.g., an exercise machine), and / or may include another system in, coupled to, and / or operated by, a computing device other than the wearable device 100 and / or the external computing device 504. For example, in some embodiments, the intelligent system 512 may constitute and / or include an intelligent audio system, an intelligent lighting system, an intelligent HVAC system, and / or an intelligent exercise system (e.g., an intelligent exercise machine). In these or other embodiments, the wearable device 100 and / or the external computing device 504 may facilitate the implementation of one or more health-enhancing features of the smart system 512, such as: health-enhancing audio features of a smart audio system; health-enhancing lighting features of a smart lighting system; health-enhancing ambient temperature features of a smart HVAC system; health-enhancing exercise features (e.g., specific exercise modes or settings) of a smart exercise system; and / or another health-enhancing feature of the smart system 512.

[0116] In some embodiments described herein, the wearable device 100 and / or external computing device 504 may send instructions to a smart system 512, which, when executed by such a system (e.g., via one or more processors), may cause the system to perform operations to implement one or more health-enhancing features of such a system. In one embodiment, the wearable device 100 and / or external computing device 504 may send instructions to a smart audio system, which, when executed by such a system (e.g., via one or more processors), may cause the system to notify the user 10 that an abnormal body temperature has been detected. In another embodiment, the wearable device 100 and / or external computing device 504 may send instructions to a smart HVAC system, which, when executed by such a system (e.g., via one or more processors), may cause the system to output air at a specific health-enhancing temperature (e.g., a specific temperature that may be defined by the user 10). In one embodiment of this disclosure, wearable device 100 and / or external computing device 504 may send instructions to a smart exercise system, which, when executed by such a system (e.g., via one or more processors), may cause the system to operate in a particular mode or setting and / or provide the user with suggestions for selecting such a mode or setting.

[0117] Example Method

[0118] Figure 6 A flowchart depicts an example method 600 for training a machine learning model to denoise biometric signals collected using mobile biometric sensors, according to an example embodiment of this disclosure. Although Figure 6 The steps performed in a particular order are depicted for illustrative and discussion purposes, but the method of this disclosure is not limited to the order or arrangement specifically shown. The various steps of method 600 may be omitted, rearranged, combined, and / or adapted in various ways without departing from the scope of this disclosure.

[0119] At position 602, the computing system can generate training data for training a machine learning biometric denoising model.

[0120] At position 604, in order to generate training data, the computing system can obtain information describing multiple clean biometric signals measured using the corresponding medical device.

[0121] At 606, in some implementations, in order to generate training data, for each of a plurality of clean biometric signals, the computing system may select at least a portion of the clean biometric signal.

[0122] At point 608, to generate training data, the computational system may sample noise from one or more noise sources among a plurality of candidate noise sources. In some implementations, before sampling noise from the plurality of candidate noise sources, the computational system may select one or more noise sources from the plurality of candidate noise sources for each of a plurality of clean biometric signals. In some implementations, selecting one or more noise sources from the plurality of candidate noise sources may include performing a randomized noise selection process to select one or more noise sources from the plurality of candidate noise sources for a clean biometric signal. In some implementations, selecting one or more noise sources from the plurality of candidate noise sources may include performing a noise selection process that evaluates the clean biometric signal to select one or more noise sources from the plurality of candidate noise sources for the clean biometric signal.

[0123] At 610, in order to generate training data, the computational system may apply noise to at least a portion of the clean biometric signal. In some implementations, the computational system may selectively apply noise to at least a portion of the clean biometric signal. In some implementations, the computational system may perform a randomized signal selection process to select at least a portion of the clean biometric signal. Alternatively or additionally, in some implementations, the computational system may perform a signal selection process that evaluates the clean biometric signal to select at least a portion of the clean biometric signal.

[0124] At position 612, the computing system can use a machine learning biometric denoising model to process the training data to obtain multiple denoised biometric signals.

[0125] At point 614, the computational system can train a machine learning biometric denoising model based on a loss function that evaluates the differences between multiple denoised biometric signals and multiple clean biometric signals. Specifically, in some implementations, the computational system can obtain information describing the multiple clean biometric signals. The multiple clean biometric signals may each include multiple clean electrocardiogram (ECG) signals measured using a medical ECG device. The computational system can process the training data using the machine learning biometric denoising model to obtain the multiple denoised ECG signals.

[0126] Training data may include multiple noisy ECG signals. The computational system can train a machine learning biometric denoising model based on a loss function. Each of the multiple noisy ECG signals may include information indicating the P wave, and the loss function can be configured to train the machine learning biometric denoising model to preserve the P wave information in each of the multiple denoised ECG signals. Alternatively, in some implementations, the computational system may obtain multiple clean PPG signals measured using a medical PPG device.

[0127] Specifically, in some implementations, the computational system can train a machine learning biometrics denoising model based on a loss function that includes one or more loss terms. Some of these loss terms can be configured to enforce certain consistency (or lack thereof) between the model's input and output. For example, the P-wave is a biologically significant wave found in ECG waveforms. However, in some cases, training a machine learning biometrics denoising model with a default loss function can cause the model to identify the P-wave as noise and then "denoise" (i.e., remove) the P-wave from the ECG waveform. To prevent this behavior, the loss function can include a consistency loss term. The consistency loss term can enforce consistency between the P-wave in the model's input and output. In other words, the consistency loss term can train the machine learning biometrics denoising model not to identify the P-wave as noise.

[0128] Alternatively, in some implementations, the loss function may include a consistency loss term for diphtheria notches in the PPG signal. A diphtheria notch refers to a feature of the PPG waveform that may be more or less pronounced when present in the waveform. Similar to the P wave, the presence / absence (and amplitude) of the diphtheria notch can have some biological significance. Therefore, the consistency loss term can enforce consistency between the diphtheria notches in the model's input and output. In other words, the consistency loss term can train a machine learning biometric denoising model to not identify diphtheria notches as noise.

[0129] In some implementations, the computing system can provide a machine learning biometric denoising model to the user's computing device for denoising biometric signals received at the user's computing device.

[0130] In some implementations, the computing system can train a machine learning biometric denoising distillation model based on a machine learning biometric denoising model. The computing system can provide the machine learning biometric denoising distillation model to the user's computing device for denoising biometric signals received at the user's computing device.

[0131] In some implementations, the computing system can receive noisy biometric signals obtained by the user's computing device. The computing system can then process the noisy biometric signals using a machine learning biometric denoising model to obtain reduced-noise biometric signals. The computing system can then send information describing the reduced-noise biometric signals back to the user's computing device.

[0132] In some implementations, the computing system can perform a biometric analysis process on the noise-reduced biometric signal to obtain biometric measurements. The computing system can send information describing the noise-reduced biometric signal to the user's computing device. This information describing the noise-reduced biometric signal may include biometric measurements.

[0133] Figure 7 A flowchart depicts an example method 700, according to an example embodiment of the present disclosure, utilizing a machine learning model for denoising biometric signals collected using mobile biometric sensors. Although Figure 7 The steps performed in a particular order are depicted for illustrative and discussion purposes, but the method of this disclosure is not limited to the order or arrangement specifically shown. The various steps of method 700 may be omitted, rearranged, combined and / or adapted in various ways without departing from the scope of this disclosure.

[0134] At point 702, a user computing device (e.g., a mobile computing device, a wearable computing device, etc.) can obtain a noisy biometric signal via a mobile biometric measurement device associated with the user computing device. The noisy biometric signal may include a first noise level. In some implementations, the user computing device may be a wearable computing device including a mobile biometric measurement device. The mobile biometric measurement device may include a surface of the wearable computing device that is in contact with the user of the wearable computing device. Obtaining the noisy biometric signal via the mobile biometric measurement device may include measuring the noisy biometric signal from the user via the surface of the wearable computing device in contact with the user using the mobile biometric measurement device.

[0135] In some implementations, obtaining noisy biometric signals via a mobile biometric measurement device associated with a user's computing device may include receiving noisy biometric signals from a wearable computing device worn by the user. The wearable computing device may include a mobile biometric measurement device. The wearable computing device may use the mobile biometric measurement device to measure noisy biometric signals from the user via a surface of the wearable computing device that contacts the user.

[0136] In some implementations, obtaining a noisy biometric signal via a mobile biometric measurement device associated with a user computing device may include obtaining a noisy biometric signal via a mobile biometric measurement device associated with a user computing device. The noisy biometric signal may include a noisy ECG signal. The mobile biometric measurement device may include a mobile ECG device. Alternatively, in some implementations, the noisy biometric signal may include a noisy PPG signal, and the mobile biometric measurement device may include a mobile PPG device.

[0137] At point 704, the user computing device can cause the noisy biometric signal to be processed by a machine learning biometric denoising model to obtain a denoised biometric signal including a second noise amount less than a first noise amount. The machine learning biometric denoising model can be trained based on a loss function that evaluates the differences between multiple noisy biometric training signals and multiple clean biometric ground truth signals. Multiple noisy biometric training signals can be generated by applying noise from randomly selected noise sources to randomly selected portions of the clean biometric signal.

[0138] In some implementations, processing noisy biometric signals using a machine learning biometric denoising model may include processing the noisy biometric signals with the machine learning biometric denoising model to obtain denoised biometric signals that include a second noise amount less than a first noise amount.

[0139] In some implementations, processing noisy biometric signals with a machine learning biometric denoising model may include processing the noisy biometric signals with a machine learning biometric denoising distillation model to obtain denoised biometric signals. The machine learning biometric denoising distillation model may be trained by a computing system via distillation from the machine learning biometric denoising model. In some implementations, the user computing device may receive the machine learning biometric denoising distillation model from the computing system before processing the noisy biometric signals with the machine learning biometric denoising distillation model.

[0140] At 706, the user computing device can provide biometric measurements to display to the user of the user computing device. The biometric measurements may be based at least in part on denoised biometric signals. In some implementations, information from the computing system indicating the denoised biometric signals may include the biometric measurements. In some implementations, providing biometric measurements to the user of the user computing device may include performing a biometric analysis process on the noise-reduced, denoised biometric signals to obtain the biometric measurements.

[0141] Additional Public Content

[0142] This paper discusses technologies related to servers, databases, software applications, and other computer-based systems, as well as the actions taken and the information sent to and from such systems. The inherent flexibility of computer-based systems allows for a wide variety of possible configurations, combinations, and partitions of tasks and functionality between and within components. For example, the processes discussed in this paper can be implemented using a single device or component, or multiple devices or components working in combination. Databases and applications can be implemented on a single system or distributed across multiple systems. Distributed components can operate sequentially or in parallel.

[0143] While the subject matter has been described in detail with respect to various specific exemplary embodiments, each example is provided by way of explanation and not limitation. Modifications, alterations, and equivalents of such embodiments will readily arise for those skilled in the art upon acquiring an understanding of the foregoing. Therefore, this disclosure does not exclude such modifications, alterations, and / or additions to the subject matter that will be readily understood by those of ordinary skill in the art. For example, features shown or described as part of an embodiment may be used with another embodiment to produce yet another embodiment. Therefore, this disclosure is intended to cover such modifications, alterations, and equivalents.

Claims

1. A computer-implemented method, comprising: Training data for training a machine learning biometric denoising model is generated by a computing system including one or more processor devices, wherein generating the training data includes: The computing system obtains information describing multiple clean biometric signals measured using the corresponding medical device; For each of the plurality of clean biosignal signals: The computing system samples noise from one or more noise sources among a plurality of candidate noise sources; and The computing system applies the noise to at least a portion of the clean biometric signal; The computing system processes the training data using the machine learning biometric denoising model to obtain multiple denoised biometric signals; and The computing system trains the machine learning biometric denoising model based on a loss function, which evaluates the difference between the plurality of denoised biometric signals and the plurality of clean biometric signals.

2. The computer-implemented method of claim 1, wherein, The method further includes: The computing system provides the machine learning biometric denoising model to the user computing device for denoising biometric signals received at the user computing device.

3. The computer-implemented method of claim 1, wherein, The method further includes: The computing system trains a machine learning biometric denoising distillation model based on the machine learning biometric denoising model; and The computing system provides the machine learning biometric denoising distillation model to the user computing device for denoising biometric signals received at the user computing device.

4. The computer-implemented method as described in claim 1, further comprising: The computing system receives noisy biometric signals obtained by the user computing device from the user computing device. The user computing device processes the noisy biometric signal using the machine learning biometric denoising model to obtain a biometric signal with reduced noise. as well as The computing system sends information describing the reduced noise level of the biometric signal to the user computing device.

5. The computer-implemented method of claim 4, wherein, The information sent to the user computing device describing the reduced noise in the biometric signal includes: The computing system performs a biometric analysis process on the noise-reduced biometric signal to obtain biometric measurements; and The computing system sends information describing the reduced noise of the biometric signal to the user computing device, wherein the information describing the reduced noise of the biometric signal includes the biometric measurement.

6. The computer-implemented method of claim 1, wherein, Before sampling the noise from one or more of the plurality of candidate noise sources, the method includes: for each of the plurality of clean biometric signals: The computing system selects one or more noise sources from the plurality of candidate noise sources.

7. The computer-implemented method as described in claim 6, wherein, Selecting one or more noise sources from the plurality of candidate noise sources includes: The computing system performs a randomized noise selection process to select one or more noise sources from the plurality of candidate noise sources for the clean biometric signal.

8. The computer-implemented method as described in claim 6, wherein, Selecting one or more noise sources from the plurality of candidate noise sources includes: The computing system performs a noise selection process to evaluate the clean biometric signal in order to select one or more noise sources from the plurality of candidate noise sources for the clean biometric signal.

9. The computer-implemented method as described in claim 1, wherein, Applying the noise to at least a portion of the clean biometric signal includes: The computing system performs a randomized signal selection process to select at least a portion of the clean biometric signal.

10. The computer-implemented method as described in claim 1, wherein, Applying the noise to at least a portion of the clean biometric signal includes: The computing system performs a signal selection process to evaluate the clean biometric signal in order to select at least a portion of the clean biometric signal.

11. The computer-implemented method as described in claim 1, wherein, Sampled noise from one or more noise sources among the plurality of candidate noise sources includes: The computing system selects one or more noise sources from the plurality of candidate noise sources, wherein the one or more noise sources include at least one of the following: Noise extracted from noisy biometric signals measured using the corresponding medical device; Including one or more noise colors from a plurality of noise colors, including Brownian noise and white noise; or One or more signal adjustment processes within a plurality of signal adjustment processes, including: The signal compression process compresses a portion of the clean biometric signal; The signal trimming process removes a portion of the clean biometric signal; or The signal expansion process stretches a portion of the clean biometric signal.

12. The computer-implemented method as described in claim 1, wherein, Obtaining the information describing the plurality of clean biometric signals measured using the corresponding medical device includes: The computing system obtains information describing the plurality of clean biometric signals, wherein the plurality of clean biometric signals respectively include a plurality of clean ECG signals measured using a medical electrocardiogram (ECG) device.

13. The computer-implemented method as described in claim 12, wherein, Processing the training data includes: The computing system processes the training data using the machine learning biometric denoising model to obtain multiple denoised ECG signals, wherein the training data includes corresponding multiple noisy ECG signals; and Training the machine learning biometric denoising model includes: The computational system trains the machine learning biometric denoising model based on a loss function, wherein each of the plurality of noisy ECG signals includes information indicating the P wave, and wherein the loss function is configured to train the machine learning biometric denoising model to retain the information indicating the P wave in each of the plurality of denoised ECG signals.

14. The computer-implemented method as described in claim 1, wherein, Obtaining the information describing the plurality of clean biometric signals measured using the corresponding medical device includes: The computing system obtains information describing the plurality of clean biometric signals, wherein the plurality of clean biometric signals respectively include a plurality of clean PPG signals measured using a medical photoplethysmography (PPG) device.

15. A user computing device, comprising: One or more processor devices; One or more tangible non-transitory computer-readable media storing computer-readable instructions that, when executed by the one or more processor devices, cause the user computing device to perform operations, including: A noisy biometric signal is obtained via a mobile biometric measurement device associated with the user computing device, wherein the noisy biometric signal includes a first noise level; The noisy biometric signal is processed using a machine learning biometric denoising model to obtain a denoised biometric signal including a second noise amount less than the first noise amount, wherein the machine learning biometric denoising model is trained based on a loss function that evaluates the difference between a plurality of noisy biometric training signals and a plurality of clean biometric ground truth signals, and wherein the plurality of noisy biometric training signals are generated by applying noise from randomly selected noise sources to randomly selected portions of the clean biometric signal; and Biometric measurements are provided to be displayed to the user of the user computing device, wherein the biometric measurements are based at least in part on the denoised biometric signals.

16. The user computing device of claim 15, wherein, The user computing device includes a wearable computing device, the wearable computing device including the mobile biometric measurement device, wherein the mobile biometric measurement device includes the wearable computing device contacting the user's surface; and The acquisition of the noisy biometric signal via the mobile biometric measurement device includes: The noisy biometric signal is measured from the user by the mobile biometric measuring device via contact with the user's surface through the wearable computing device.

17. The user computing device as claimed in claim 15, wherein, Obtaining the noisy biometric signal via the mobile biometric measurement device associated with the user computing device includes: The noisy biometric signal is received from a wearable computing device worn by the user, wherein the wearable computing device includes the mobile biometric measuring device, and wherein the wearable computing device measures the noisy biometric signal from the user via the mobile biometric measuring device through contact between the wearable computing device and the user's surface.

18. The user computing device as claimed in claim 15, wherein, The process of processing the noisy biometric signal using the machine learning biometric denoising model to obtain the denoised biometric signal includes: The noisy biometric signal is processed using the machine learning biometric denoising model to obtain the denoised biometric signal, which includes a second noise level less than the first noise level.

19. The user computing device as claimed in claim 18, wherein, Processing the noisy biometric signal using the machine learning biometric denoising model includes: The noisy biometric signal is processed using a machine learning biometric denoising distillation model to obtain the denoised biometric signal, wherein the machine learning biometric denoising distillation model is trained by a computational system via distillation from the machine learning biometric denoising model; and The method includes the following steps before processing the noisy biometric signal using the machine learning biometric denoising distillation model: The machine learning biometric denoising distillation model is received from the computing system.

20. The user computing device as claimed in claim 15, wherein, The process of processing the noisy biometric signal using the machine learning biometric denoising model to obtain the denoised biometric signal includes: Send the noisy biometric signal to the computing system for processing at the computing system using the machine learning biometric denoising model; and In response to transmitting the noisy biometric signal, information indicating the denoised biometric signal is received from the computing system.

21. The user computing device of claim 20, wherein, The information from the computing system indicating the denoised biometric signal includes the biometric measurements.

22. The user computing device as claimed in claim 15, wherein, The user who provides the biometric measurements to be displayed to the user computing device includes: A biometric analysis process is performed on the denoised biometric signal with reduced noise to obtain the biometric measurement.

23. The user computing device as claimed in claim 15, wherein, Obtaining the noisy biometric signal via the mobile biometric measurement device associated with the user computing device includes: The noisy biometric signal is obtained via the mobile biometric measurement device associated with the user computing device, wherein the noisy biometric signal includes a noisy ECG signal, and wherein the mobile biometric measurement device includes a mobile ECG device.

24. The user computing device as claimed in claim 15, wherein, Obtaining the noisy biometric signal via the mobile biometric measurement device associated with the user computing device includes: The noisy biometric signal is obtained via the mobile biometric measurement device associated with the user computing device, wherein the noisy biometric signal includes a noisy PPG signal, and wherein the mobile biometric measurement device includes a mobile PPG device.

25. One or more tangible non-transitory computer-readable media storing computer-readable instructions that, when executed by one or more processor devices, cause the one or more processor devices to perform operations, the operations including: Generating training data for training a machine learning biometric denoising model, wherein generating the training data includes: Obtain information describing multiple clean biometric signals measured using the corresponding medical device; and For each of the plurality of clean biosignal signals: Select at least a portion of the clean biometric signals; Sample noise from one or more noise sources from a plurality of candidate noise sources; and The noise is applied to at least a portion of the clean biometric signal; The training data is processed using the machine learning biometric denoising model to obtain multiple denoised biometric signals; and The machine learning biometric denoising model is trained based on a loss function that evaluates the difference between the plurality of denoised biometric signals and the plurality of clean biometric signals.

Images