Systems and methods involving sleep management

Abstract

[Solution] An example embodiment is directed to a system including a processing circuit and a memory circuit storing a predictive data model that indicates different patterns and probabilities of a user transitioning from a wakeful state to a sleep state. The processing circuit uses data indicative of the user's current psychophysiological state to detect, from the different patterns in the predictive data model, a pattern that indicates a probability that the user will transition from a wakeful state to a sleep state at a certain date and time. The processing circuit further selects, based on the detected pattern, an intervention action that is predicted to increase the probability that the user will transition from a wakeful state to a sleep state at a certain date and time, and communicates a message to the user indicating the intervention action.

Description

[Technical Field]

[0001] Sleep is necessary for the health and well-being of individuals. Sleep deprivation is widespread in modern society. Approximately one-third of the U.S. population has sleep problems, and this proportion is increasing. Psychological stress is recognized as a factor related to sleep deprivation. Furthermore, people living in modern society are often stressed, worried, and troubled, which leads to psychophysiological activation before sleep (hyperarousal), making it difficult to fall asleep and achieve restful sleep. [Background technology]

[0002] The present disclosure is directed to solving these and other problems related to a user's sleep management.

[0003] Various embodiments of the present disclosure monitor and / or manipulate factors that affect sleep and transition times to improve sleep onset (e.g., smoother physiological changes in the transition from wakefulness to sleep), improve sleep processes (e.g., lower arousal threshold, increase sleep efficiency, enhance nocturnal cardiovascular function), and / or shorten transition times, often referred to as "wake-to-sleep transition times" (e.g., faster perceived and / or physiological sleep onset latency). In some embodiments, the systems, devices, and / or methods may additionally or alternatively be used to manage acute states of exaggerated psychophysiological activation and / or manipulate acute psychophysiological conditions.

[0004] Particular embodiments are directed to a system including a processing circuit and a memory circuit. The memory circuit stores a predictive data model indicating different patterns and probabilities of a user's transition from a wakeful state to a sleep state. The processing circuit uses data indicating the user's current psychophysiological state to detect, from the different patterns in the predictive data model, a pattern that indicates a probability that the user will transition from the wakeful state to a sleep state at a certain date and time. The processing circuit further selects an intervention action based on the detected pattern that is predicted to increase the probability that the user will transition from the wakeful state to a sleep state at a certain date and time, and communicates a message to the user indicating the intervention action. The increased probability that the user will transition from the wakeful state to a sleep state at a certain date and time optionally includes an increased probability of reducing the user's time to transition to a sleep state compared to if the intervention action were not taken.

[0005] In some embodiments, the processing circuit detects patterns in the data by identifying a Feature Set from among a plurality of Feature Sets and using the Feature Set to select sub-models of the predictive data model. The predictive data model includes a plurality of sub-models indicative of the probability of the user transitioning to a sleep state in response to different intervention actions, the plurality of sub-models being associated with a particular Feature Set of the plurality of Feature Sets. Each feature of a particular Feature Set and / or each of the plurality of Feature Sets may have a weight associated with the probability of the user transitioning to a sleep state. In some embodiments, the plurality of sub-models are associated with different time frames.

[0006] In some embodiments, the processing circuitry modifies the predictive data model based on feedback data indicating whether the user transitions to a sleep state in response to an intervention action.

[0007] In some embodiments, the processing circuitry receives the feedback data in real time and communicates another message indicating a modified intervention action in response to the feedback data and the modified predictive data model. For example, the processing circuitry may receive the feedback data and, in response to the received feedback data, identify features of the feedback data, identify whether the user will respond to the intervention action predicted by the predictive data model to increase the probability based on the identified features, and, in response to an unexpected response, modify the predictive data model for the user associated with the detected pattern. The modified intervention action may be for a currently occurring sleep session (e.g., real-time adjustment) and / or for future sleep sessions.

[0008] In some embodiments, the intervention is part of a sleep intervention strategy that includes multiple interventions, which may be selected from behavioral interventions, cognitive interventions, neuromodulatory interventions, environmental modifications, sensory interventions, and combinations thereof.

[0009] In some embodiments, the processing circuit communicates a message indicating a sleep intervention strategy and including a sequence of multiple intervention actions.

[0010] In some embodiments, the processing circuit communicates a plurality of messages including a message indicating a plurality of intervention actions, each of the plurality of messages may be selected from the group consisting of a message to a user instructing the user to take the respective intervention action and a message to another device to automatically initiate the respective intervention action at a specific time according to the sleep intervention strategy.

[0011] In some embodiments, the system further comprises an input circuit for receiving data indicative of the user's current psychophysiological state, which may comprise a wearable physiological sensor for sensing the user's physiological signals and another sensor for sensing an atmospheric measurement.

[0012] In some embodiments, the input circuitry receives data indicative of a current psychophysiological state of the user, the received data being selected from the group consisting of schedule or calendar data, stress levels, general mood, dietary data, health information, exercise data, sleep data, and combinations thereof.

[0013] Various example embodiments are directed to a non-transitory recording medium comprising instructions that, when executed, cause a processing circuit to identify a feature set from a plurality of feature sets of data indicative of a user's current psychophysiology; detect a pattern associated with a predictive data model indicative of the user's likelihood of transitioning from a wakefulness state to a sleep state at a certain date and time based on the identified feature set; communicate a message to the user indicative of an intervention action that is predicted to increase the user's likelihood of transitioning from a wakefulness state to a sleep state at a certain date and time based on the detected pattern and the predictive data model; and modify the predictive data model based on feedback data indicative of the user's transitioning to a sleep state in response to the intervention action.

[0014] In some embodiments, the instructions for detecting a pattern include executable instructions to select a sub-model of the associated predictive data model using the identified feature set. The instructions may be further executed to communicate another message in response to the feedback data and the modified predictive data model indicating a modified intervention action.

[0015] In some embodiments, the instructions for modifying the predictive data model include instructions executable to modify a weight associated with an intervention action in response to the feature set identified for the user. The weight for the intervention action may relate to a probability that the intervention action will improve sleep and / or reduce time to sleep. For example, the intervention action may improve the probability of sleep onset and / or sleep processes.

[0016] In some embodiments, the instructions to modify the predictive data model include instructions executable to modify the predictive data model for the user over time based on the feedback data and additionally received feedback data indicative of different sleep intervention strategies and respective multiple feature sets.

[0017] In some embodiments, the intervention action is part of a sleep intervention strategy that includes multiple intervention actions. For example, the instructions may be executable to communicate a message indicating the sleep intervention strategy and including a sequence of the multiple intervention actions, and to modify the predictive data model, including modifying the sequence of the multiple intervention actions and one or more of the multiple intervention actions.

[0018] Various embodiments are directed to a system comprising an input circuit, a memory circuit, and a processing circuit. The input circuit receives data indicative of a user's current psychophysiological state. The memory circuit stores a predictive data model indicative of different patterns and probabilities of the user transitioning from a wakeful state to a sleep state. The processing circuit uses the data to detect, from the different patterns in the predictive data model, patterns indicative of a probability that the user will transition from a wakeful state to a sleep state at a certain date and time, identifies, based on the detected patterns, a sleep intervention strategy including at least one intervention action predicted to increase the probability that the user will transition from a wakeful state to a sleep state at a certain date and time, and communicates a message to the user indicative of the at least one intervention action.

[0019] In some embodiments, the memory circuitry includes instructions that, when executed, cause the memory circuitry to generate a predictive data model based on general population trends and publicly available data, and to modify the predictive data model for a user over time using feedback data indicating the success of different sleep intervention strategies for each feature set.

[0020] In some embodiments, the at least one message indicating the at least one intervention action further includes an indication of the order and timing of the at least one intervention action. For example, the processing circuitry may communicate at least one message to another device to automatically take the at least one intervention action at a specific time in accordance with the sleep intervention strategy.

[0021] Embodiments according to the present disclosure include all combinations of the described embodiments. Further embodiments and the full scope of applicability of the invention will become apparent from the detailed description provided below. However, it will be understood that the detailed description and specific examples, while indicating preferred embodiments of the invention, are given by way of illustration only, since the spirit and scope of the invention will be apparent to those skilled in the art from this detailed description. All publications, patents, and patent applications cited herein are hereby incorporated by reference in their entirety for all purposes, including any citations therein.

[0022] Various example embodiments may be more fully understood in consideration of the following detailed description in conjunction with the accompanying drawings. [Brief explanation of the drawings]

[0023] [Figure 1] FIG. 1 illustrates an example of a system for sleep management, according to various embodiments. [Figure 2] FIG. 2 illustrates an example computing device with a non-transitory storage medium containing executable instructions according to the present disclosure. [Figure 3] FIG. 3 illustrates another example of a system for sleep management, according to various embodiments. [Figure 4] FIG. 4 illustrates another example of a sleep management system according to various embodiments. [Figure 5] FIG. 5 illustrates an example method of data processing by the sleep management system of FIG. 4, according to various embodiments. [Figure 6] FIG. 6 illustrates an example of a sleep intervention strategy, according to various embodiments. [Figure 7]FIG. 7 illustrates an example process for generating a predictive data model, according to various embodiments. [Figure 8] 8A-C illustrate different examples of predictive data models, according to various embodiments. [Figure 9] FIG. 9 illustrates sample effects of interventions on heart rate (HR), heart rate variability (HRV), and use of an intervention planner to generate a sleep intervention strategy for a user, according to various embodiments. [Figure 10] FIG. 10 illustrates inter-beat interval times during a person's sleep onset with and without intervention (in this case, virtual reality biofeedback), according to various embodiments. [Figure 11] 11A and 11B illustrate the group-level relationship between perceptions of cognitive alertness before sleep and subsequent physiological time spent falling asleep for individuals with and without insomnia, according to various embodiments. [Figure 12] FIG. 12 illustrates the relationship between physiological pre-sleep activation state (cortisol levels) and nighttime polysomnographic sleep efficiency, according to various embodiments. [Figure 13] FIG. 13 shows a diagram of pre-sleep effects on sleep, according to various embodiments. [Figure 14] FIG. 14 shows a theoretical plot of the relationship between pre-sleep physiological autonomic activation (eg, HRV) and subsequent nighttime sleep efficiency, according to various embodiments.

[0024] While the various embodiments described herein are susceptible to modifications and alternative forms, such embodiments have been shown by way of example in the drawings and will be described in detail. It will be understood, however, that the intention is not to limit the invention to the particular embodiments. Rather, the invention covers all modifications, equivalents, and alternatives within the scope of this disclosure, including embodiments defined in the claims. Furthermore, the term "embodiment" as used throughout this application is for illustrative purposes only and is not limiting. DETAILED DESCRIPTION OF THE INVENTION

[0025] Aspects of the present disclosure are believed to be applicable to a variety of systems and methods, including sleep management systems that track sleep states over time and can identify which biological, personal, or environmental factors make it difficult for a user to transition from a wakeful state to a sleep state. In certain embodiments, the system may communicate to the user interventions that increase the likelihood of transitioning to a sleep state, and the interventions are updated over time based on feedback data indicating the success or failure of the interventions. The present disclosure is not necessarily limited to such applications, and various aspects of the present disclosure will be understood through the description of various examples using this context.

[0026] Thus, in the following description, various specific details are set forth to describe the particular embodiments presented herein. However, it will be apparent to one of ordinary skill in the art that other embodiments and / or variations of one or more of these embodiments may be implemented without all of the following detailed description. In other embodiments, well-known features are not described in detail so as not to obscure the description of the embodiments herein. For ease of illustration, identical elements or additional instances of identical elements in different figures are designated by the same reference numerals.

[0027] Embodiments of the present disclosure include a system for improving a user's sleep by optimizing the user's psychophysiological state at bedtime or other sleep attempt times (e.g., during a nap). Such embodiments are adaptive and personalized to the user. Embodiments are directed to combating hyperarousal (e.g., altered states of psychophysiological activation) that disrupt a user's restful night's sleep. The system uses data indicative of the user's current psychophysiological state to detect patterns indicative of the user's likelihood of transitioning from a wakeful state to a sleep state at a given time and date, and selects, based on the detected patterns, an intervention action that is predicted to increase the user's likelihood of transitioning from a wakeful state to a sleep state at a given time and date. The intervention action includes or relates to a sleep intervention strategy that increases or improves the user's sleep. Improving sleep may include improving sleep onset (smoother psychophysiological changes in the transition from wakefulness to sleep), improving sleep processes (reducing arousal threshold, improving sleep efficiency, enhancing nighttime cardiovascular function), and / or shortening the time it takes to transition from wakefulness to sleep (faster perceived / organized sleep onset latency). In some embodiments, the sleep intervention strategy is a pre-sleep intervention strategy. In various embodiments, the strategy may span the entire wake-to-sleep transition, not limited to pre-sleep. The system communicates a message to the user indicating the intervention action, either directly to the user or to another device that performs or communicates with the user.

[0028] In some embodiments, the predictive data model may be associated with multiple different patterns. Each pattern may include or be associated with a different feature set of factors, the order and timing of occurrence of the factors, among other characteristics of the factors, such as amplitude and intensity, that may affect the user's psychophysiological state. In some embodiments, the factors may affect the time it takes for the user to transition from a wakeful state to a sleep state, also referred to herein as "transition time." The patterns indicate different probabilities that the user will transition to a sleep state at a particular date and time, and / or different intervention actions that can improve the probability and / or transition time. In some embodiments, different patterns are associated with different feature sets, and the detected patterns are used to select sub-models of the predictive data model. For example, data may be input into the predictive data model, and the input data may be used to identify patterns (e.g., feature sets of multiple feature sets) and output the user's probability of transitioning to a sleep state. In some embodiments, the output of the predictive data model includes an intervention action that improves the probability of sleep and / or the transition time to sleep. Additionally or alternatively, the intervention action may be used to manipulate the user's acute psychophysiological state, such as improving relaxation or reducing sympathetic activation. In another embodiment, the output is used to select an intervention action.

[0029] "Hyperarousal" herein includes or refers to abnormalities in psychophysiological activity that are reflected across different biological (e.g., elevated heart rate, high cortisol levels, elevated high-frequency EEG activity), psychological (e.g., worry, anxiety, rumination), and emotional domains and may be the result of an exaggerated response to stimuli (e.g., hyperarousal may be triggered by stress, fear of not being able to sleep, and / or financial concerns). Hyperarousal may be used to describe a condition that worsens in a user when the user is not relaxed or anxious, such that the transition from a wakefulness state to a sleep state is prolonged or hindered.

[0030] As used herein, "wakefulness" includes or refers to a state in which a person remains conscious and aware of their surroundings. Biologically, wakefulness may be characterized by asynchronous, rapid, and low-amplitude EEG activity. As used herein, "sleep" includes or refers to a state in which a person is less responsive to environmental stimuli and less aware of their surroundings. Biologically, sleep involves a gradual increase in cortical synchronization and a general state of physiological deactivation (e.g., decreased metabolism).

[0031] Embodiments of the present disclosure are directed to targeting factors that prevent a user from falling asleep and getting a good night's sleep, such as factors that can cause hyperarousal at bedtime. Hyperarousal can manifest as a racing heart, muscle tension, and worries about sleep, as well as other non-sleep-related worries such as mental overactivity, strategic thinking, pain-related discomfort, and anxiety. Hyperarousal can manifest differently in different people and have different causes, involving various domains as noted above.

[0032] In some embodiments, different intervention behavioral strategies can be used to target specific aspects of hyperarousal. Different users often have unique sleep and relaxation needs, and these needs, as well as factors that affect stress and sleep, change over time. For example, users have different personalities, respond differently to external stimuli, and live in different socioeconomic environments, which may require individualized sleep promotion methods. Embodiments of the present disclosure are directed to predictive data models that are used to recognize that users have different preferences for relaxation techniques and sleep methods, and that the same methods will work differently for different users.

[0033] The predictive data model can be adaptive and personalized to support user relaxation and sleep by focusing on the transition from wakefulness to sleep. For example, the predictive data model can be used to shorten the transition time from wakefulness to sleep by reducing wakefulness before sleep. Various embodiments target psychophysiological states throughout the sleep onset process, which is the time window a user uses to reach a specific psychophysiological state that enables the transition from wakefulness to sleep. The predictive data model can adapt to users and over time by identifying different factors that contribute to psychophysiological arousal at bedtime (e.g., psychophysiological stress, consumption of stimulants or other substances, excessive physical activity in the evening, pain) and recognizing that the factors vary within and between users. The predictive data model can recognize that different interventions or multiple interventions targeting pre-sleep hyperarousal (e.g., cognitive-behavioral interventions, relaxation strategies, virtual reality immersion, neuromodulation, brainwave entrainment) have different efficacy in reducing pre-sleep hyperarousal, and that efficacy varies between users and over time. For example, users have different preferences for how to relax, and these preferences can interfere with the effectiveness of interventions. Furthermore, different strategies or combinations have different instantaneous effects (e.g., minute-by-minute resolution), and the effects can vary over different days, months, etc.

[0034] In various embodiments, the system dynamically senses data indicative of physiological (e.g., HR via plethysmography sensors) and behavioral (e.g., amount of daily physical activity), environmental conditions (e.g., humidity, temperature, light), and provides multi-component (e.g., cognitive and behavioral interventions, meditation, virtual reality immersion, direct current stimulation, soothing sounds) interventions (e.g., speakers, virtual reality headsets, drug patches) via one or more actuators to reduce stress and promote and sustain sleep based on the user's sleep and relaxation needs. The system includes a predictive data model, which may include machine learning (ML) and / or artificial intelligence (AI) techniques, to optimize interventions for the user in real time, taking into account multiple features such as the user's psychophysiological state and the effectiveness of specific sleep intervention strategies at any particular time. The system and predictive data model: 1) targeting factors such as hyperarousal during time frames such as the sleep onset phase, often referred to herein as the "transition time" for falling asleep and maintaining sleep, and 2) implementing AI and / or ML to arrive at a highly adaptive and personalized intervention of one or more components that maintains efficacy and user engagement over time.

[0035] The system and methods for implementing the system can be timed, specific, adaptable, and personalized to enable optimization of the user's psychophysiological state during the pre-sleep period. Due to the diversity and constant dynamic modification of characteristics contributing to sleep deprivation and stress, user preferences, and the effectiveness of different treatment methods for users, the system can tailor intervention actions for effectiveness, generalizability, and sustainability. Furthermore, the timing and specificity of intervention actions can be tailored.

[0036] Specific embodiments are directed to a system for managing and / or improving sleep. In some embodiments, the system is a personalized sleep management system that helps minimize a user's transition time from wakefulness to sleep. However, embodiments are not so limited, and the system can improve sleep in various ways, such as minimizing transition time, improving a user's ability to fall asleep, and improving their ability to stay asleep. The system collects data about the user, analyzes the collected data, and selects an intervention action to minimize the user's transition time from wakefulness to sleep based on the data and a predictive data model. The system can collect data about the user, including data about which mitigation intervention strategies have been most successful in previous similar cases, and present a strategy plan to the user. In some embodiments, the user can provide feedback to the system about which interventions worked under a predetermined set of conditions. The sleep management system can provide external interventions based not only on data input but also by drawing on previous intervention sessions to arrive at a sleep intervention strategy for the user's current psychophysiological state prior to sleep.

[0037] Various embodiments are directed to a system comprising input circuitry (e.g., a sensor or other type of circuitry), a memory circuit, and a processing circuit. The input circuitry obtains current physical measurements associated with a user and can use the physical measurements, by themselves or in combination with other data, to provide an intervention strategy for the user for the day. The physical measurements obtained by the input circuitry can include one or more physiological signals, environmental information, etc. In some embodiments, the wearable device is associated with a sleep management system, and can track the user's state throughout the day and help devise a plan to minimize the user's wake-to-sleep transition time.

[0038] In some embodiments, the processing circuit generates a predictive data model for minimizing transition times between wakefulness and sleep states for an individual user. The processing circuit can update and modify the proposed sleep intervention strategy based on daily tasks, stressors, physiological measurements, and user input. User input may include mood, consumed items, planned interactions with others, etc. In some cases, the processing circuit can synchronize with the user's daily calendar as a starting point for building a sleep intervention strategy for when the user is ready to sleep. The processing circuit can take multiple intervention actions based on the input data, sensed or measured data, and data obtained via ML. For example, the sleep management system may advise the user that a meditation session before attempting to sleep may be helpful based on the user's calendar for that day. As another example, the sleep management system may advise the user to drink hot tea an hour and a half before attempting to sleep based on the user's dietary input for that day.

[0039] 1 illustrates an example of a system for sleep management, according to various embodiments. System 100 can be used to manage a user's 108 transition from a wakefulness state to a sleep state. For example, system 100 can optimize the psychophysiological state of user 108 over a time window used for the sleep onset process, i.e., for user 108 to reach a particular psychophysiological state to enable the transition from wakefulness to sleep and / or improve sleep.

[0040] The system 100 includes a processing circuit 102 and a memory circuit 104. In some embodiments, the processing circuit 102 and the memory circuit 104 form part of the device 105, such as by being part of the logic of the device 105. In other embodiments, the processing circuit 102 and the memory circuit 104 form part of separate devices that communicate with each other, such as a distributed computing device over a network. The memory circuit 104 can store a predictive data model 106. In some embodiments, the memory circuit 104 further stores instructions executable by the processing circuit 102 to cause the processes described below. The predictive data model 106 can indicate different patterns and probabilities of a user 108 transitioning from a wakefulness state to a sleep state. In some embodiments, the predictive data model 106 can be associated with intervention actions to increase the probability of transition and / or reduce the transition time at a date and time, as further described herein. In some embodiments, different patterns include or are associated with different feature sets. For example, the patterns may include a user's sleep patterns associated with different feature sets. The different patterns may include: The predictive data model 106 and feature set may indicate different probabilities that the user 108 will transition from a wakefulness state to a sleep state at a given date and time, and / or may indicate different intervention actions that can increase the probability of the transition and / or decrease the transition time.

[0041] As used herein, features include factors that may affect the psychophysiological state of a user 108, the order and / or timing of factors, the current time and / or date, and other characteristics of factors. For example, such factors may affect the sleep of a particular user 108 and / or multiple users (e.g., generally, based on demographics, etc.). As further described herein, some factors may negatively affect sleep, such as negatively affecting (e.g., increasing) the transition time between a user's awake and sleep states. Some factors may positively affect sleep, such as positively affecting (e.g., shortening) the transition time. For example, a feature set may relate to factors that prevent a person from falling asleep, such as causing hyperarousal around the time the user goes to bed. Hyperarousal can manifest as not only racing heartbeats, muscle tension, stress, or worry, but also other symptoms such as mental overactivity, strategic thinking, discomfort related to pain, and anxiety. Hyperarousal manifests differently in different people, has different causes (e.g., contributing factors), and involves various domains (e.g., cognition, emotion, physiology).

[0042] In some embodiments, the predictive data model 106 may comprise multiple sub-models. The multiple sub-models are associated with multiple patterns (e.g., feature sets). Each pattern may be associated with at least one sub-model. In some embodiments, multiple sub-models may be associated with each pattern, such that each pattern is associated with multiple sub-models. As an example, each of the multiple sub-models for each pattern may be associated with a different time frame (e.g., milliseconds, seconds, minutes, hours, days, weeks, months), and the multiple sub-models may be run simultaneously. Each of the multiple sub-models may indicate the probability of a user transitioning from a wakefulness state to a sleep state at a particular time of day and / or at different transition times in response to different intervention actions.

[0043] Each sub-model may be associated with a particular Feature Set of the plurality of Feature Sets, and each feature in a particular Feature Set may have a weight. The weight may be based on or indicate how predictive each feature is for a user to transition or not transition into a sleep state and / or the feature's impact on transition time. In some embodiments, each intervention action may differentially increase the probability of transition into a sleep state and / or differentially decrease transition time for different sub-models and associated Feature Sets (e.g., factors affecting sleep for a time frame). As previously described, the predictive data model 106 may be used to predict transitions into sleep states based on intervention actions and improve the user's 108 sleep cycles and transition times over time.

[0044] Processing circuitry 102 may use data indicative of the current psychophysiological state of user 108 to detect different patterns in predictive data model 106 that are indicative of the probability that user 108 will transition from a wakefulness state to a sleep state at a given date and time. Processing circuitry 102 may process the data to identify a feature set from the received data and use the feature set to detect the pattern of features.

[0045] The predictive data model 106 may include an AI model or a machine learning model (MLM). Various ML frameworks are available from multiple providers that offer open-source ML datasets and tools that enable developers to design, train, validate, and deploy MLMs, such as AI / ML processors. AI / ML processors (often referred to as hardware accelerators (MLAs) or neural processing units (NPUs)) can accelerate ML processing. ML processors are integrated circuits (ASICs) that can be designed with multi-cores and employ precision processing with dataflow architectures and optimized memory usage to speed up calculations and improve computational throughput when processing MLMs.

[0046] For example, the predictive data model 106 may receive input data, identify patterns from the input data, and output a probability that the user will transition to a sleep state based on the pattern. In some embodiments, the output may be used to select an intervention action to increase the probability of transition to a sleep state and / or decrease the time to transition to a sleep state. In some embodiments, the output includes the intervention action. For example, the output may include a probability that the user will transition to a sleep state in response to the intervention action.

[0047] Examples of data indicative of a user's current psychophysiological state include factors such as physiological measurements, daily patterns such as daily routines, consumption data, environmental data, and reproductive information, among other data. For example, physiological measurements include, but are not limited to, HR, HRV, respiratory rate, muscle tone, body temperature, forehead temperature, and skin conductance. Other examples of data include circadian rhythm, the user's reproductive stage (e.g., puberty, menstrual cycle, menopause), personality traits (e.g., introversion, narcissism, depression), medical conditions, and demographics (e.g., age, gender, race, type of work), among other data. Environmental data may include light intensity, noise, temperature, humidity level, and pollen level, among other environmental factors. Consumption data may include consumption of food, liquids, caffeine, and other drugs (e.g., nicotine, prescription drugs, medical cannabis, non-prescription drugs and substances), the amount consumed, and the time of consumption. Exercise data may include the type, duration, and / or intensity of exercise. Other user behaviors or routines may also be input, such as sexual activity, relaxation activities (e.g., meditation, reading), television or other media viewed (e.g., reading on a computer, watching movies), and stressful activities (e.g., job interviews, traveling, driving in public transportation, political events, social events). In some embodiments, relaxing and stressful activities may be identified by the user 108 (e.g., manually labeled by the user) and / or using the predictive data model 106. For example, data input may include user preferences, such as identified intervention behaviors (e.g., guided meditation vs. breath awareness, yoga, spiritual vs. non-spiritual, and type of music).

[0048] Data can be input from a variety of sources. In some embodiments, sources may include other devices, such as sensor circuitry, third-party platforms, mobile applications, and web platforms. Data can be general to a user population or specific to a user. For example, data general to a user population may include information based on clinical studies, recommendations, and journals. Such data can be used to generate predictive data models 106 and / or as feedback data. For example, scientific data may include associations of scientifically informed features and targeted intervention plans that may result from scientific studies, including clinical trials. Data can be provided automatically or manually entered by the user 108. For example, calendar data, geolocation data, perceived alertness, exercise data, consumption data, emotional state, and user preferences can be obtained by other applications, sensors, and / or self-entered by the user 108 into a user interface (UI).

[0049] In some embodiments, processing circuit 102 can detect patterns by identifying a feature set from multiple feature sets from the input data and selecting a sub-model from multiple sub-models of predictive data model 106 based on the identified feature set. As described above, the multiple sub-models indicate the probability that user 108 will transition to a sleep state in response to different intervention actions and may be associated with a particular feature set from the multiple feature sets. In some embodiments, the multiple sub-models include, for each pattern, a subset of sub-models associated with different time frames, including short, medium, and long time frames, and each feature in the set may have a weight associated with the probability that the user will transition to a sleep state and / or an impact on the transition time (e.g., associating different feature sets with time frames).

[0050] Based on the detected patterns, the processing circuit 102 may select an intervention action predicted to improve sleep, such as increasing the probability that the user will transition to a sleep state at a certain date and time and / or improving sleep onset and / or sleep processes. The selection may include an output of the predictive data model 106 or may be based on the predictive data model 106. The intervention action may be part of a sleep intervention strategy that includes an intervention action and / or multiple intervention actions. Examples of intervention actions include behavioral intervention actions, cognitive intervention actions, neuromodulatory actions, environmental changes, sensory actions, and combinations thereof. The intervention action may be passively instructed by the user 108 to perform or may be actively instructed by another device to perform (e.g., play music, turn down the lights, lower the temperature, etc.). Examples of intervention actions are described further herein. The intervention actions may be used to target different hyperarousal states.

[0051] Processing circuit 102 may communicate a message indicating an intervention action to user 108. The message may be communicated directly to user 108, such as via a UI (user interface) associated with processing circuit 102. In another embodiment, processing circuit 102 may communicate a message to another device to perform an intervention action or to communicate a message to user 108.

[0052] In some embodiments, processing circuit 102 may communicate a message or messages related to multiple intervention actions, including the intervention action. For example, processing circuit 102 may communicate a message indicating a sleep intervention strategy and including a sequence of the multiple intervention actions. In some embodiments, processing circuit 102 may communicate multiple messages, including a message indicating the multiple intervention actions, each selected from the group consisting of a message displayed to a user instructing the user to take each intervention action, and a message to another device to automatically generate each intervention action at a specific time according to the sleep intervention strategy.

[0053] Intervention actions and / or sleep intervention strategies may be different for different users. For example, different strategies can be used to target specific aspects of hyperarousal. Each user may have different sleep and relaxation needs, as well as factors that affect sleep and may change over time for a particular user. Also, different users may have different personalities, be in different environments, and respond differently to different intervention actions (e.g., external stimuli). For example, a particular user may have different preferences for relaxation techniques and sleep methods, and similar methods may have different effects. As described further herein, the predictive data model 106 can be informed of such preferences and / or learn them over time using feedback data.

[0054] In some embodiments, the processing circuit 102 generates the predictive data model 106 and stores the predictive data model 106 in the memory circuit 104. In another example, the processing circuit 102 receives the predictive data model 106 and stores the predictive data model 106 in the memory circuit 104. As previously mentioned, the predictive data model 106 may include an MLM that is trained using input from a dataset.

[0055] Generating the predictive data model 106 may include receiving input data including factors influencing sleep state transitions for the user 108 or multiple users (e.g., used to form a feature set), receiving known outputs related to the input, identifying distinct patterns indicative of sleep state transitions (or non-transitions) within the input data, and, based on the patterns, identifying a predictive probability that the user 108 will transition to a sleep state at a given date and time using the additionally received input data. The known outputs may include past sleep transition times, time periods, and other information. In some embodiments, the known outputs may include intervention actions that can be generated to increase the probability of transitioning to a sleep state and / or decrease transition times and that can be used to identify future intervention actions. In some embodiments, the intervention actions may otherwise or additionally improve the user's sleep, such as by providing a smoother physiological transition from wakefulness to sleep, reducing arousal thresholds, increasing sleep efficiency, and / or enhancing nighttime cardiovascular function. The processing circuit 102 receives the inputs and known outputs and uses them to generate the predictive data model 106. The inputs and known outputs may be actively input, passively input, or received. The inputs and known outputs may include lifestyle data including, but not limited to, reported sleep cycles, schedule or calendar data, stress levels, mood, exercise data, sleep data, and dietary data, and health information including, but not limited to, physiological signals or parameters, medications, diagnoses, and other treatments, and various combinations thereof.

[0056] The predictive data model 106 may be dynamically updated over time. For example, the processing circuit 102 may modify the predictive data model 106 based on feedback data indicating whether the user 108 transitions to a sleep state in response to an intervention action. Such feedback data may include an identification of whether or not a sleep state transition occurs at a particular date and time, the transition time, and characteristics occurring during the transition (such as characteristics occurring when the user transitions and / or the intervention action). Other feedback data may include, among other data, sensor data indicating physical measurements from the user 108, the user's self-report of sleep quality, etc. The processing circuit 102 may receive the feedback data and, in response to the received feedback data, identify characteristics from the feedback data and, based on the identified characteristics, identify whether the user 108 will respond to the intervention action predicted by the predictive data model 106 to increase the probability. In response to an unexpected response, the processing circuit 102 may modify the predictive data model 106 for the user 108 to relate to the detected pattern.

[0057] In some embodiments, processing circuit 102 may receive feedback data in real time. In response to the feedback data and the modified predictive data model 106, processing circuit 102 may communicate another message indicating modified intervention actions. In this manner, system 100 can adaptively adjust a sleep intervention strategy for user 108 in real time. For example, the adjusted sleep intervention strategy may include adjusted intervention actions applied during a currently occurring sleep session (e.g., adjusted in real time) and / or for future sleep sessions.

[0058] In some embodiments, feedback data is input by the user 108 and identifies user reports of changes in sleep cycles and / or sleep quality over time. The updating may include adjusting the weighting of different features based on experienced sleep patterns and / or sleep state transitions. As a specific example, over time, the user 108 may experience changes in their sleep cycles that make transitions to sleep states more difficult due to different features. Alternatively and / or additionally, the feedback data may indicate specific information about the current sleep state. It is understood that different users may experience a reduction in and / or increased ease of sleep state transitions with different intervention actions. The system 100 can learn which sleep intervention strategy, such as an intervention action or series of intervention actions, is most effective at a given time and / or in response to different feature sets to relax the user 108 and improve their sleep. Because the strategy depends on the current feature set, the system 100 can provide the most effective strategy on a day-to-day basis. The system 100 uses data from multiple sources, at individual or population-based levels, and on different time scales.

[0059] As a specific example, system 100 may learn that whenever user 108 has a meeting (according to their calendar) scheduled for early the next morning and their current pre-sleep arousal level exceeds a certain threshold (e.g., a resting HR 6 bpm above baseline), the best sleep intervention strategy is 5 minutes of respiratory biofeedback followed by 10 minutes of progressive muscle relaxation. Over time, system 100 may learn that under the same circumstances, if user 108 does not exercise that evening (information from a third-party platform, such as a wearable activity tracker, or information from another wearable device that has the disclosed system built into or partially integrated into the wearable device), then 5 minutes of respiratory biofeedback is not necessary and 10 minutes of progressive muscle relaxation will lead to the same level of effectiveness in system 100 outcomes.

[0060] As previously described, processing circuit 102 can communicate data indicative of an intervention action. For example, processing circuit 102 may communicate a message to user 108 to take an intervention action to increase the probability and / or otherwise induce a transition to a sleep state. The intervention action may be based on the user's past response to the intervention action by system 100 and / or other user responses, such as feedback data. In other embodiments and / or in addition, the intervention action includes computer-readable instructions communicated to other devices, such as sensor circuitry, temperature control circuitry (e.g., associated with a heating, ventilation, and air conditioning (HVAC) system), and / or light control circuitry. As non-limiting examples, the intervention action may be an instruction provided to an HVAC system to change the temperature in a particular room, an instruction to a user device, such as ringing a smartwatch to notify the user to turn off the device or perform other actions, and / or a display on an application executed by a smartphone instructing the user on a particular action to take, among other specific actions. It is understood that embodiments are not limited to one action, and multiple intervention actions may be initiated by system 100. Other example intervention actions may include user suggestions or stress-relief actions that can increase the probability of transitioning to a sleep state or improve the user's sleep. Example stress-relief strategies may include cognitive behavioral therapy, music therapy (e.g., having the user activate or recommending a device that plays music), hormone therapy and / or aromatherapy (e.g., activating or recommending a device that outputs a scent, such as lavender, chamomile, or rose scent), etc. Example suggestions may include recommending that the user exercise, take supplements, engage in aromatherapy or hormone therapy, play music, reduce caffeine intake, chew gum, and practice deep breathing.

[0061] As previously mentioned, system 100 may further include input circuitry. The input circuitry may receive data indicative of the user's current psychophysiological state. The data may include schedule or calendar data, stress levels, general mood, dietary data, health information, exercise data, sleep data, and combinations thereof. In some embodiments, the input circuitry may include sensor circuitry used to obtain physical measurements related to the user 108. In some embodiments, the input circuitry includes wearable physiological sensors for sensing physiological signals from the user 108 and sensors for sensing air measurements. An input circuit, such as a sensor circuitry, includes communication circuitry for communicating the physical measurements to processing circuit 102. The communication circuitry may communicate wirelessly or via wires.

[0062] The physical measurements may be physiological signals or measurements (e.g., bodily fluids) from the user 108, movement, and / or atmospheric measurements from the environment surrounding or near the user 108. For example, the input circuitry may include a wearable physiological sensor, such as a wearable device, that senses physiological signals from the user. Alternatively or additionally, the input circuitry may include a sensor that senses atmospheric measurements. Examples of physiological signals include parameters such as blood pressure, HR, skin conductance, body temperature, etc. Examples of atmospheric measurements include atmospheric temperature, atmospheric pressure, humidity, etc. Embodiments are not limited to physiological signals or atmospheric measurements and may additionally or alternatively include movement data (e.g., from an accelerometer) and / or global positioning data (GPS).

[0063] Thus, in some embodiments, the system 100 may comprise, but is not limited to, a UI (e.g., an application, a wearable device, a web platform), sensors (e.g., GPS sensors, accelerometers, microphones, skin conductance, photoplethysmography, ambient light, skin and ambient temperature sensors, pressure and chemical sensors, microphones), actuators (e.g., speakers, screens, virtual reality headsets, neuromodulators, drug patches, lights) for providing intervention actions (e.g., cognitive behavioral interventions, meditation strategies, neuromodulation, biofeedback, virtual immersion, light illumination, brainwave entrainment), connectivity modules (e.g., WiFi, Bluetooth), a cloud computing system, and an AI-based data processing module.

[0064] System 100 includes evidence-based, science-based ML and AI-based methods. System 100 can increase and / or maximize the effectiveness of single or multi-component interventions to alleviate stress across the wake-to-sleep transition and improve sleep. For example, system 100 can provide personalized intervention packages that are dynamically optimized based on user needs, the user's physiological state, and other factors over time and in response to changing individual needs.

[0065] The system 100 can learn which sleep intervention strategy (one intervention action or combination of intervention actions) at any given time is most effective in relaxing and improving sleep for the user 108. For example, each night at bedtime, the system 100 can provide the most effective strategy to the user 108.

[0066] System 100 is not limited to providing sleep intervention strategies focused directly on optimizing the pre-sleep psychophysiology of user 108. System 100 may promote other user behaviors, such as sleep hygiene and / or other behaviors of user 108, as well as other strategies spanning the transition from outside of bedtime and / or full wakefulness to sleep (e.g., enhancing physical activity routines, meditation, promoting a positive attitude toward life).

[0067] 2 illustrates an example of a computing device including a non-transitory storage medium containing executable instructions according to the present disclosure. The computing device according to the example of this specification includes a user device having logic circuitry such as the processing circuitry and memory circuitry illustrated in FIGS. 1-2.

[0068] The computing device includes a processing circuit 220 and a memory circuit. The memory circuit includes a computer storage medium 22 that stores a set of executable instructions 224, 226, 228, 229. The computer storage medium 222 may include, for example, read-only memory (ROM), random-access memory (RAM), electrically erasable programmable read-only memory (EEPROM), flash memory, a solid-state drive, and / or a set of discrete data registers.

[0069] At 224, processing circuit 220 identifies a feature set from the plurality of feature sets from the data, the data being indicative of the user's current psychophysiological state. Processing circuit 220 may receive data indicative of the feature set.

[0070] At 226, processing circuitry 220 detects a pattern associated with a predictive data model indicative of a probability that a user will transition from a wakeful state to a sleep state at a certain date and time based on the identified feature set. Detecting the pattern may include instructions executable to identify a feature set from the plurality of feature sets in the data and select a sub-model of the predictive data model using the identified feature set.

[0071] At 228, processing circuitry 220 communicates a message to the user indicating an intervention action that is predicted to increase the likelihood that the user will transition to a sleep state at a certain date and time based on the detected pattern and the predictive data model. At 229, processing circuitry 220 modifies the predictive data model based on feedback data indicating whether the user transitioned to a sleep state in response to the intervention action and / or other information related to sleep, such as transition time.

[0072] In some embodiments, the instructions are further executed to communicate another message indicating a modified intervention action in response to the feedback data and the modified predictive data model. For example, processing circuitry 220 may modify weights provided to intervention actions in response to an identified feature set for the user in the predictive data model. In some embodiments, processing circuitry 220 can modify the predictive data model over time for the user based on the feedback data and additional received feedback data indicative of different sleep intervention strategies and feature sets (e.g., indicative of the success or failure of strategies responsive to different feature sets for the user and / or generic data provided from a database).

[0073] In some embodiments, the intervention action is part of a sleep intervention strategy that includes multiple intervention actions. In some such embodiments, processing circuit 220 may communicate a message indicating the sleep intervention strategy and the sequencing of the multiple intervention actions, and modify the predictive data model, including modifying the sequencing of the multiple intervention actions and / or the sequencing of the multiple intervention actions.

[0074] 3 illustrates another example system for sleep management, according to various embodiments. System 310 includes processing circuit 302 and memory circuit 304, which includes predictive data model 306, as described above in connection with FIG. 1, and further includes input circuit 312. In some embodiments, processing circuit 302 and memory circuit 304 form part of device 316, although embodiments are not so limited. In various embodiments, the system is for acute psychophysiological condition and sleep management.

[0075] The input circuit 312, as described above, can receive data indicative of the current psychophysiological state of the user 308. The input circuit 312 may include multiple different types of devices. As described above, the input circuit 312 may include a sensor circuit that obtains physical measurements from the user 308 and communicates the physical measurements to the processing circuit 302 using the communication circuit 314. Although the input circuit 312 is illustrated as a single circuit, embodiments are not so limited. For example, the input circuit 312 may include multiple sensors that obtain different physical measurements, such as different physiological signals, air measurements, movement data, and / or GPS signals. In particular embodiments, the input circuit 312 includes a wearable physiological sensor that senses physiological signals from the user 308 and another sensor that senses air measurements, as described above. Furthermore, the input circuit 312 is not limited to sensor circuitry and may include other circuits, such as a UI through which the user 308 inputs data.

[0076] The memory circuitry 304 stores a predictive data model 306 that indicates different patterns and probabilities of a user 308 transitioning from a wakefulness state to a sleep state, and in some embodiments, computer-readable instructions that are executable by the processing circuitry 302 to:

[0077] The processing circuit 302 can use the data to detect distinct patterns in the predictive data model 306 that are indicative of the likelihood that the user 308 will transition from a wakeful state to a sleep state at a certain date and time. As described above, the processing circuit 302 may identify a feature set from the received data and process the data to detect patterns using the feature set. Based on the detected pattern, the processing circuit 302 may identify a sleep intervention strategy that includes at least one intervention action predicted to increase the likelihood that the user 308 will transition to a sleep state at a date and time and / or otherwise improve sleep. The processing circuit 302 further communicates at least one message to the user 308 indicating the at least one intervention action.

[0078] In some embodiments, processing circuit 302 generates a predictive data model 306 that indicates the probability that a user will transition to a sleep state at a given date and time. Predictive data model 306 may be generated based on general population trends and publicly available information, among other data. In some embodiments, processing circuit 302 may modify predictive data model 306 for user 308 over time using feedback data that indicates the success and / or failure of different sleep intervention strategies for each feature set.

[0079] In some embodiments, the intervention action may include multiple actions. For example, processing circuit 302 may communicate at least one message indicating the at least one intervention action including instructions for the order and timing of the at least one intervention action. In some embodiments, processing circuit 302 may communicate at least one message to another device to automatically generate the at least one intervention action at a specific time according to the sleep intervention strategy.

[0080] The sleep management systems 100, 310 shown in Figures 1-3 can be configured in a number of ways. Some embodiments of the sleep management systems 100, 310 may be, or form part of, a computer program or application that can run on a smartphone, tablet, desk computer, laptop, smartwatch, activity tracker, or other standalone device. Other embodiments of the sleep management system may be a web-based program. In any of the above embodiments, the program may include executable instructions shown in Figure 2.

[0081] Generally, the user of a sleep management system is a single user. In other configurations of sleep management systems, multiple users may use a single system to monitor and store each user's sleep patterns.

[0082] The systems 100, 310 described herein may employ several interventions or multiple interventions to meet a user's sleep and / or relaxation needs and relaxation strategy preferences. The systems 100, 310 may use or include multiple data sources. For example, sources may include the user's physiological state, habits, behaviors, and environmental conditions with different time frames (e.g., daytime behaviors known to affect sleep, such as physical activity or caffeine consumption; real-time physiological indicators, such as current heart rate measurements; and morning self-reported measurements, such as perceived alertness) used to drive the selection and application of sleep / stress interventions. Furthermore, the systems may include multiple personalized closed loops at different time scales. This approach may be optimized by targeting a combination of different areas of hyperarousal.

[0083] The timing of intervention behaviors can be tailored to target the sleep onset process and take into account daytime behaviors known to affect sleep (e.g., amount of physical activity, caffeine consumption).

[0084] The system 100, 310 can assist users with sleep and relaxation needs (e.g., business people, military personnel, individuals with insomnia, hospitalized patients). The system 100, 310 can cover different applications, such as assisting nighttime sleep, promoting daytime naps, or simply helping people relax on a general level. Targeting stress and sleep is a prerequisite for health and well-being. User adoption of the technology, self-reported and physiological data demonstrating the system's population effectiveness, clinical trial results, scientific recognition, consumer reports, and clinical acceptance of the technology's use are among the measures that make it possible to quantify the impact of the intervention.

[0085] 4 illustrates another example sleep management system, according to various embodiments. More specifically, FIG. 4 is a functional block diagram of a sleep management system 440 that assists a user in managing their sleep cycle.

[0086] 4, sleep management system 440 includes a user module 441 that includes a data input / output (I / O) module 444, a data processing module 445, a decision module 443, and an intervention action module 442. Embodiments are not so limited, and sleep management system 440 may include fewer or more modules than shown in FIGURE 4. Each of modules 442, 443, 444, 445 includes computer-executable instructions stored on one or more non-transitory computer storage media and executable by processing circuitry, such as a single computing device or distributed across multiple computing devices.

[0087] As described further below, the sleep and relaxation of the user 450 may be promoted by the decision module 443 receiving processed information from the I / O module 444 and informed by scientific data 446 (e.g., scientific evidence regarding factors affecting psychophysiological states and sleep) to activate the intervention action module 442. The decision module 443 may receive feedback data from the I / O module 444 (e.g., psychophysiological states, transition times, sleep quality after the intervention).

[0088] In some embodiments, data from one or more users may be stored in a cloud computing system 449, which may interact with the user module 441 to optimize system performance (e.g., improve system effectiveness based on information collected about results stratified by demographics such as user age and gender, geography, etc.). The user module 441 may run independently of the data contained within the cloud computing system 449, with parameters set based on local storage since the last update of the data stored within the cloud computing system 449.

[0089] The user module 441 may incorporate online learning capabilities to learn and derive predictive data models between input data and intervention actions to be taken. In some embodiments, data may not be stored on the user's 450 device. In some embodiments, the user's 450 device may act as a local cache of the most recent user-based optimized parameters. In some embodiments, all data may be uploaded and stored in the cloud computing system 449. This data can be used to find the most appropriate selection of input features that most effectively impact the user's 450 sleep, use batch learning functionality to derive a data model between the input data and appropriate intervention actions to be taken, and frequently update the predictive data model 460 according to the cloud-derived model. The cloud computing system 449 can use data from a population of users 448 to build predictive data models to use as initial data models for new users.

[0090] The I / O module 444 allows the user 450 and / or input circuitry to interact with other modules integrated into and / or forming part of the sleep management system 440. The I / O module 444 can obtain input data from the user 450 and other devices. The data may be obtained by other modules. In some embodiments, the user 450 may actively and / or passively input information into the sleep management system 440 in various ways. The user 450 may manually enter data. The user 450 may verbally enter information into the sleep management system 440. The user 450 may upload health-related files to the sleep management system 440. In other embodiments and / or in addition, the user 450 can grant the sleep management system 440 access to data from other applications, such as calendar data, exercise or food tracking applications, and / or other external locations.

[0091] Using the I / O module 444, the user 450 can input personal data via input hardware (e.g., a mouse, keyboard, touchscreen, microphone, etc.), such as, but not limited to, demographics, body mass index (BMI), ethnicity, age, the user's reproductive stage, medications, mood at certain times of day, anxiety level, activity level, allergies, types of foods consumed, etc. The I / O module 444 may also be used to provide feedback data, such as the success or failure of an intervention. In general, the I / O module 444 can be used to obtain data from and provide communications to the user 450. The system 440 can elicit a response or prompt the user 450 to perform a particular action by displaying a message and requesting a response to be entered via the UI. Responses may not be entered in some embodiments. For example, the system 440 may include a voice processing module for interpreting speech and analyzing responses.

[0092] The I / O module 444 may receive input from various sources 463, 464, 465, 453, 446. Multiple sources may include a user 450, where data is obtained via a UI 463, sensor circuits 464, historical data 465, scientific data 446, and other data sources 453.

[0093] The data processing module 445 can process the data from the I / O module 444. Data processing may include, but is not limited to, low-pass filtering, noise removal, feature extraction, etc. The extracted features are temporal and / or spectral characteristics that represent the data and its specific time patterns, variability, and frequency content. As shown in FIG. 4, after pre-processing of the input data, features are obtained from each set of input categories at 466 and used to form a final input feature set at 468, which may be further dimensionally reduced at 469.

[0094] The raw physiological and environmental signals from the sensor circuit 464 and other sources over a preceding time window may form an input feature set. Temporal and / or spectral features, which represent signal patterns in time and frequency, may also be extracted from the raw physiological and environmental signals. Temporal features may include statistical measures such as the mean, variance, and higher-order statistics of the input data over the time window. Spectral features may be extracted using a Fourier transform. In some embodiments, spectral features may be spectral moments, spectral power fractions, spectral power peaks, and spectral power ratios. Features may also be extracted after applying an appropriate transform to facilitate understanding of the input patterns, such as wavelet analysis. Features may include factors of a model that best represents the data over a particular time window.

[0095] From the measured physiological signals, different physiologically relevant features can be extracted, such as HR, pulse rate, pulse transit time, etc.

[0096] User actions and various events can be extracted not only from the UI but also from calendars and other third-party platforms. The events may be clustered / classified into several categories to form an input feature set. Each event and user action is represented by 0s and 1s over time, where 1 indicates the event occurrence and 0 indicates otherwise. Natural language processing (NLP) techniques may be incorporated to read events from calendars and other third-party platforms and cluster / classify the events.

[0097] Each individual characteristic (this information, as well as information on quality of life, coping strategies, cognitive and emotional functioning, can be collected through a smartphone application-based questionnaire) and demographic information may form an element of an input feature set.

[0098] The extracted features may form a high-dimensional feature vector that is provided to the decision module 443. To reduce the complexity of the decision process and enhance the learning process, the dimensionality of the feature input vector can be reduced 469 using subspace learning and dimensionality reduction functions. Linear decomposition methods such as factor analysis, principal component analysis, singular value decomposition, and independent component analysis may be used. Because the input data may exhibit a high degree of nonlinearity, nonlinear dimensionality reduction methods such as kernel-based methods can also be used. Online versions of such statistical techniques can be used on the user module 441, while batch versions can be incorporated on cloud computing systems.

[0099] As more data becomes available, deep learning techniques such as autoencoder neural network configurations can be used to reduce the dimensionality of the features.

[0100] Depending on the feature selection method, this step can be performed on the cloud computing system 449, or on both the cloud computing system 449 and the user module 441. In some embodiments, different sets of features are used to train the predictive data model 460 and select the feature set with the best performance. Because this is computationally expensive and time-consuming, this approach may be performed by the cloud computing system 449. The predictive data model 460 is then downloaded onto the user software / application for further use. A strategy can be used to select the most appropriate features that can be performed on both the user module 441 and the cloud computing system 449. Other embodiments include the use of a least absolute shrinkage and selection operator (LASSO) technique, in which any features with non-zero regression coefficients are selected.

[0101] The decision module 443 receives data from the data processing module 445 and is capable of performing various types of analysis, including, but not limited to, detecting patterns in the input data and selecting intervention actions based on the detected patterns. The decision module 443 may analyze environmental conditions (e.g., ambient temperature, humidity, season, time of day, dietary structure, caffeine and / or alcohol consumption, medication use), personal conditions (e.g., mood, stress, anxiety, time of day, exercise, menstrual cycle patterns, calendar events), location via GPS and / or user input (e.g., supermarket, home, work), and physiological conditions (e.g., skin temperature, thermal sensitivity, cardiac autonomic state such as HR, HRB, skin blood flow such as peripheral vasoconstriction / vasodilation).

[0102] In various embodiments, the decision module 443 may be involved in multiple data processing streams: a first data processing stream 452 may relate to processing input data to determine the current psychophysiological state and recommended intervention actions, and a second data processing stream 454 may relate to processing feedback data from the implementation (e.g., success or failure) of the intervention actions.

[0103] An example of the first data processing stream 452 is described below. In some embodiments, the input to the decision module 443 is a reduced set of features that best represent the input data. The output of the decision module 443 is weights (or scores) for all possible intervention actions. Each weight may be provided between 0 and 1. These weights are processed by the intervention planning sub-module 458 to select an appropriate intervention action for the sleep intervention strategy. As shown in FIG. 5 , the sleep management system can collect input data. The input data may be a variety of different types of data, as described above. Once input, relevant features and correlations between the input data and the effect on wakefulness-to-sleep transition time are obtained, allowing the decision module 443 to determine the efficiency of various possible intervention actions and sequences. The decision module 443 then selects the optimal sleep intervention strategy and presents the sleep intervention strategy to the user 450 using the intervention planning sub-module 458. The sleep management system 440 can update the sleep intervention strategy throughout the day as it receives feedback data and / or new input data.

[0104] The decision module 443 may include a predictive data model 460, which includes a user-specific data model 461 and an ML model or function 462 that updates the user-specific data model 461 over time as new input-output relationships become apparent using feedback data and / or other modifications to the predictive data model 460. The user-specific data model 461 may be updated based on input data and output measurements of the system 440 through various modalities, including a UI 463, sensor circuits 464, etc. Data from the I / O module 444 may be processed to score interventions taken by the intervention module 442 on a scale of 0 to 10, among other scoring ranges. The user 450 may directly score the interventions on a scale of 0 to 10, where 0 indicates dissatisfaction and 10 indicates satisfaction. Different information may be extracted from the sensor circuits 464 and used in accordance with a scientific framework identified from the scientific data 446 to score the interventions taken on a scale of 0 to 10. For example, changes in HR or respiration rate extracted from the sensor circuitry 464 may be compared before and after an action is taken and used to generate a score from 0 to 10 for each action according to a scientific framework.

[0105] The following describes an example of the second data processing stream 454. In some embodiments, the decision module 443 uses the scientific data 446 to process the feedback data and generate relevant information. For example, HR and respiration rate are extracted from the sensor circuit 464 before and after an action is taken. The scores of the intervention actions are added together and scaled between 0 and 1, which is used as the target value for ML.

[0106] A simple example of a predictive data model 460 is a regression model. Another example is a neural network such as a multi-layer perceptron (MLP). Such models can be trained on the fly (user module 441) as new data becomes available, or offline as data is uploaded to the cloud computing system 449. For example, incremental learning techniques can be used to train the network on the fly, and batch techniques can be used on the cloud computing system 449.

[0107] The cloud computing system 449 may include a module similar to the user module 441 that can use data received from all users 448 and build a general data model that can be incorporated as an initial predictive data model for new users without triggering the intervention action module 442. The predictive data model 460 is then updated individually as the user continues to use it. The cloud computing system 449 also builds an individual data model for each user based on all data received from the user over time. Batch learning methods can be used to derive the models.

[0108] Various approaches can be used to address missing input data. For example, a user may not always accurately report the time and amount of substance intake, meals, physical activity, or other input. As another example, a user may not always wear a sensor device or devices that collect physiological signals, and / or overall data input may be fragmented to include a subset of inputs necessary to reliably be used by the decision module 443 to trigger the intervention action module 442. In some embodiments, missing data may be replaced by a constant, a random value, or the average value of available input samples. Other techniques, such as interpolation or predictive modeling, such as regression or hidden Markov models, may be applied. Another approach is to omit the input feature set containing missing data. Additionally, some variations of subspace ML, designed to handle missing data issues, such as principal component analysis, may be used.

[0109] As mentioned above, the inputs may include continuous and / or discrete information sources of different timescales (e.g., real-time user biosignals, characteristic information such as personality). Examples of inputs include psychophysiological states (e.g., HR and HRV, respiratory rate, muscle tone, forehead temperature, body temperature, skin conductance), circadian rhythms, reproductive stages (e.g., puberty, menstrual cycle stage, menopause), personality traits (e.g., narcissism, introversion), environment (e.g., light intensity, noise, temperature), user behaviors (e.g., time and amount of caffeine intake, time and intensity of exercise, food intake, sexual activity, drug and substance use), stressful events (e.g., job interviews, international travel, driving, political and social events), demographics (e.g., age, race, gender), user preferences (e.g., guided meditation vs. breath awareness, yoga, spiritual vs. non-spiritual, and type of music), etc. The inputs may be measured using sensors and / or obtained from other sources, such as third-party platforms, mobile applications, and web platforms.

[0110] In some embodiments, the data processing module 445 can process data in the form of self-reports from the user, physiological and environmental signals, and / or as provided by third-party platforms. The decision module 443 can integrate data inputs, as well as data outputs and feedback data from the cloud computing system 449 and from the scientific framework to trigger the intervention action module 442.

[0111] In some embodiments, the input includes or is based on scientific data 446. The scientific data 446, along with other data from the I / O module 444, provides information to the decision module 443. For example, the data may be based on results from scientific publications about factors that affect a user's stress and sleep, and the effectiveness of intervention modules to reduce stress and improve a user's sleep. Non-limiting examples of wake-promoting and sleep-promoting factors are provided below in Table 1. [Table 1]

[0112] The intervention action module 442 may include a set of intervention actions that may be triggered by the decision module 443. The decision module 443 may control the order and duration of the intervention actions. The intervention actions may be performed singly or simultaneously. Examples of intervention actions include behavioral interventions, cognitive interventions, environmental changes, neuromodulation, and sensory stimulation. Examples of behavioral interventions include respiratory relaxation, respiratory awareness, progressive muscle relaxation, body scans, respiratory biofeedback, guided meditation, and mindfulness circles, among other interventions. Examples of cognitive interventions include guided cognitive exercises, identifying distorted thoughts, cost-benefit analysis, and detective thinking, among other interventions. Examples of environmental changes or interventions include adjusting ambient lighting, humidity, and temperature, among other interventions. An example of a neuromodulation intervention includes transcutaneous vagus nerve stimulation. Examples of sensory interventions include brainwave entertainment, such as playing audio, turning off a television or other device, binaural beats, and binaural audio. The user may select preferred actions and / or combinations for falling asleep, and the attributes of the actions or combinations (duration, sequence, etc.) may be determined by the results of the decision module 443.

[0113] The intervention behaviors may be provided by the intervention behavior module 442 using various delivery channels and / or I / O modules 444. Examples of delivery channels include a mobile application 456, actuators 451 (e.g., audio, video, sensory stimuli), and a web platform 457. Outputs may include different sources of information, continuous and / or discrete, at different time scales. Different types of outputs may include the user's psychophysiological state (e.g., HR, HRV, respiratory rate, muscle tone, forehead temperature), social interactions (e.g., meeting friends, time spent on social media), mood (e.g., anxiety, aggression, lethargy), and sleep (e.g., time to fall asleep, sleep quality, sleep duration), etc. Output data may be measured using the sensor circuitry 464 (e.g., physiological changes), third-party platforms (e.g., weight using a weight management application), mobile applications (e.g., self-reported assessments of sleep, emotions, and cognition, cognitive test results), and web platforms (e.g., self-reported assessments, interviews with experts), etc.

[0114] The decision module 443 may constantly process feedback data from the / O module 444. For example, if a "guided meditation" intervention is active and no physiological changes in the user (e.g., a decrease in HR) are detected, it may determine that the current intervention is not physiologically relaxing for the user 450. The decision module 443 may then execute a different intervention (e.g., respiratory relaxation).

[0115] The data output can be processed in a variety of ways. The cloud computing system 449 may store system information for single users and multiple users 448. This data allows for stratification of the effectiveness of the sleep management system according to the individual.

[0116] As previously mentioned, system 440 may receive data not only from a user-entered data but also from other sources, including, but not limited to, physiological sensors, environmental sensors, data entered into other applications running on the same platform, etc. With respect to physiological sensors, sensor circuitry 464 may include, but is not limited to, skin conductance sensors, skin temperature sensors, blood pressure sensors, pulse sensors, photoplethysmography sensors, electrocardiogram sensors, and electroencephalogram sensors. Other physiological sensors may sense biological signals, such as for sweat analysis. Sensors of various physical forms may be utilized, including, but not limited to, sensors adhesively coupled to the body, sensors housed in a wearable, sensors coupled to or attached to clothing, etc.

[0117] With regard to environmental sensors, the system 440 may receive data from external sensors capable of communicating with the system 440 or from sensors already integrated within the system. In an example of the system 440 existing in the form of a mobile application, sensors already present in the mobile device may be utilized to provide input data to the system. Environmental data may include time of day, local temperature, local humidity, light exposure, etc. Some of this data may be measured locally if sensors are utilized, but in some cases, data may be obtained from another application or external source, such as a website. In this case, the system may automatically initiate a search to obtain such data regarding the user's location. Such data input may be conveniently obtained when the system exists in the form of a mobile application. With regard to obtaining data from other applications, such data may include, but is not limited to, meeting data from a calendar application or data from other menstrual management applications that may be obtained through interaction with one or more of these applications.

[0118] As previously mentioned, the data processing module 445 processes the data to make it more suitable for analysis. Data from each source can be processed differently. For example, processing of data from physiological sensors may include, but is not limited to, low-pass filtering, averaging, smoothing, etc. Microphone data for emotional analysis may also be similarly filtered to remove artifacts. Furthermore, data from websites may be used directly. Sensors that generate complex waveforms, such as electrocardiogram (ECG) and electroencephalogram (EEG) sensors, may also be plotted using extracted parameters. In the case of ECG, the data module 443 may extract numerous parameters, such as pulse-return (PR) interval, QRS complex duration, and RR interval, and store this information along with the date and time of the measured parameters.

[0119] In various embodiments, the sleep management system 440 may include or form part of a computer program or application executable on a smartphone, tablet, desk computer, laptop, smartwatch, activity tracker, or other separate device that may be included within or in data communication with a wearable device.

[0120] FIG. 5 illustrates an example method of data processing by the sleep management system of FIG. 4 , according to various embodiments. As previously described, input data is collected at 571, and features are extracted from the input data at 572, such as by using a data processing module. The extracted features are input at 573 to a decision process module, which may include a predictive data model. The predictive data model is used to determine the effectiveness of different intervention actions at 574, and a sleep intervention strategy is generated at 575. The intervention actions of the intervention strategy are presented to the user at 576. In response, feedback data, such as data indicative of the user's physiological response, may be collected at 577. Relevant features are extracted from the feedback data at 578. It is determined at 579 whether a physiological response is expected. If so, the intervention plan or strategy is continued at 576, physiological response data is collected at 577, and features are extracted at 578. If a physiological response is not expected, it is determined at 580 whether sleep or relaxation has been achieved. If so, the intervention strategy is terminated at 581. If not, at 573, feedback regarding the failure of the intervention strategy is sent back to the data processing module to modify the predictive data model.

[0121] FIG. 6 illustrates an example sleep intervention strategy, according to various embodiments. The following is an example of sleep onset during periods of above-threshold physical activity: It is 9:46 PM, and the user desires to sleep. Input data includes: i) real-time environmental noise indicates a low level of noise, similar to the previous three days; ii) the user has a two-hour meeting at 8 AM the next day (from the user's mobile phone calendar); iii) the user self-reports physical discomfort but no anxiety (user input); iv) the user's real-time HR level is approximately 3 beats per minute (bpm) higher than the user's pre-sleep baseline HR (average over the past three days); and v) the user engaged in aerobic exercise between 7 AM and 8 AM (from a wearable device linked to the user's smartphone). The decision process module can then execute an intervention action targeting physiological arousal (e.g., respiratory relaxation).

[0122] For example, Figure 6 shows an example of intervention actions performed by the decision process module based on several levels of information (data input, feedback data, scientific data). Intervention actions may be performed serially (sequentially) and / or in parallel (simultaneously). Intervention actions 1, 2, 4, 5, and 6 may be performed at different times and durations. In a specific example, intervention action 1 is respiratory relaxation (Type = Behavioral, Appropriate 1 = , Appropriate 2 = , etc.), intervention action 2 is progressive muscle relaxation (Type = Behavioral, Appropriate 1 = , Appropriate 2 = , etc.), intervention action 3 is body scan (Type = Behavioral, Appropriate 1 = , Appropriate 2 = , etc.), intervention action 4 is transcutaneous vagus nerve stimulation (Type = Neuromodulation, Appropriate 1 = , Appropriate 2 = , etc.), intervention action 5 is relaxing binaural audio (Type = Sensory Stimulation, Appropriate 1 = , Appropriate 2 = , etc.), and intervention action 6 is respiratory biofeedback (Type = Behavioral, Appropriate 1 = , Appropriate 2 = , etc.). The top graph 683 shows the intervention actions and different times, and the bottom graph 685 shows different real-time physiological output data at different times. Different intervention actions may be selected based on the resulting decrease in the user's HR thereafter. In a particular example, intervention action 5 may be implemented based on the user's preference for a relaxing audio background. Intervention actions 1 and 2 may not be successful, but intervention action 6 may be successful in reducing physiological activation.

[0123] 7 illustrates an example process for generating a predictive model, according to various embodiments. As previously described, the determination module 443 illustrated by FIG. 4 may be implemented to use and / or generate a predictive data model 792 that indicates the probability that a user will transition to a sleep state at a given date and time based on input data. In some embodiments, the predictive data model 792 outputs data that indicates intervention actions that are predicted to improve or cause the probability.

[0124] The predictive data model 792 receives multiple different input data, such as: 1) physiological signals collected by non-invasive sensors and systems; 2) user routines, including data extracted from the user's calendar, GPS location, etc.; 3) environmental sensors, including temperature, humidity, etc.; and 4) user self-reports or inputs to the system, such as mood, energy levels or physical state, food or drink intake over the course of a day, medications, supplements, etc. The predictive data model 792 may also receive information from health websites, online feeds, and / or scientific findings in the literature, such as journal articles, which may serve as additional inputs. The output of the predictive data model 792 is the current and / or future probability that the user will transition into a sleep state, which may be in response to an intervention action. In some embodiments, the probability is based on or increases based on the occurrence of the intervention action. In a non-limiting example, the intervention action module may be programmed to communicate a message to a smart HVAC system to lower the temperature or to a smartphone to play music.

[0125] The inputs are processed to obtain an indication that allows the decision process module to learn patterns of feature sets related to sleep state transitions or non-transitions for a particular user, which may include using different ML processes to generate sub-models 790-1, 790-2, 790-3, 790-4. Each of the sub-models 790-1, 790-2, 790-3, 790-4 may be associated with a different time frame for the feature sets represented by input categories 1-4 and / or associated with a different feature set and / or output sleep intervention strategy. Different ML processes may be incorporated into the predictive data model 792 depending on the input category of the data. ML may be used to construct the sub-models 790-1, 790-2, 790-3, 790-4 between the inputs and outputs that are current and / or future probabilities of sleep transition occurrence. Based on laboratory "gold standard" measurements of sleep occurrence (e.g., polysomnography) and / or non-laboratory accepted sleep measurements (e.g., actigraphy-based sleep / wake displays) and user self-reported input about sleep, the features of the predictive data model 792 are optimized by minimizing a cost function for each of the sub-models 790-1, 790-2, 790-3, 790-4. The cost function is a function that maps the model prediction probability to a real number that intuitively represents some "cost" associated with the predicted probability value.

[0126] Below is an example of input data: This example is not limiting and additional or fewer categories may be used.

[0127] As shown in FIG. 7, an example of the first input data category may include raw physiological signals or their extracted features. The extracted features may be temporal and / or spectral features that represent the physiological signals and their specific temporal patterns, variance, and frequency content. Temporal features may include statistical measures such as the mean, variance, and higher-order statistics of the input data over a time window. Spectral features may be extracted using a Fourier transform. Examples of spectral features may be spectral moments, spectral power fractions, spectral power peaks, and spectral power. Features may also be extracted after applying an appropriate transform, such as a wavelet transform, that facilitates understanding of the input patterns. The features may include parameters of a model that best represents the data in a specific time window. Due to the very high dimensionality of the input patterns, statistical techniques such as principal component analysis and linear component analysis can be used to transform the features into a lower-dimensional subspace to achieve a more accurate and efficient representation of the input patterns.

[0128] A second category of input data may include inputs related to the user's routine at a particular time, such as a 24-hour period, and outputs the probability of transitioning to a sleep state at a particular time.

[0129] For this purpose, ML techniques such as multiple regression, genetic programming, support vector regression, and differential neural network structures may be used. As an example, for this purpose, an ML neural network with m outputs may be trained. However, the output layer may consist of m nodes with logistic activation functions, such that each output is between 0 and 1. The cost function to be minimized may be the mean squared error, and the optimization algorithm may be set to backpropagation.

[0130] The different events in the calendar can be extracted using NLP techniques and classified or clustered into several groups according to their similarities. As an example, algorithms such as centroid classification, Naive Bayes, etc. can be used to classify the different events into predefined classes. In another example, partitioning algorithms such as K-Means, or hierarchical algorithms including agglomerative and partitioning approaches can be used to cluster the events.

[0131] Input from a GPS can be used to form a similar matrix of time-location inputs. If there is a relationship between calendar events and GPS locations and the occurrence of sleep transitions, a 3D data chart can be created in which elements represent specific events at specific locations and specific times and can be represented by coordinates such as shown in Figure 7.

[0132] An example of a third input category may include environmental or atmospheric data such as temperature, humidity, etc., which may be treated the same as category 1. In some embodiments, the ML process may be designed in a way that outputs probabilities for some future time interval, e.g., every 30 minutes. As an example, a neural network may be employed. The output layer may consist of multiple outputs, each representing a probability for a particular time interval in the future. The remaining ML, optimization, and cost functions remain the same.

[0133] An example of a fourth input category includes the user's mood, which may be obtained each morning as part of a self-report, for example, by scoring from 0 to 10. An MLP, where the output activation function is a linear function, may be used to model this.

[0134] The predictive data model 792 provides the probability that a sleep state transition will occur at a certain date and time based on observed and / or collected data and / or in response to a sleep intervention strategy. A simple example is a logistic regression model that defines a linear decision boundary between training samples that are associated with sleep transitions and those that are not. If there are more complex or non-linear relationships between inputs and outputs, more complex models can be constructed. In such cases, deep neural networks such as MLPs may be used.

[0135] To calculate optimal network parameters (weights and biases in the case of neural networks), an optimization operation such as backpropagation can be performed, which can be done in a batch or incremental manner: in a batch manner, the network is fed with all available training data to calculate the optimal parameters, or in an incremental manner, the parameters are updated each time a training sample is presented to the network.

[0136] Optimization of model parameters can be done by minimizing a cost function. An example of a cost function is the cross-entropy error that the system defines between the estimated probabilities and the "true" sleep distribution. Given a dataset of N training samples, the cross-entropy cost function is: It is defined as JPEG0007753255000002.jpg20152. Here, t for the training sample i is the true sleep probability, which is either 0 or 1, and y i is the predicted probability and can be any value between 0 and 1.

[0137] During training, the cost function is minimized by adjusting the model parameters so that inputs corresponding to sleep occurrences have output probabilities close to 1, and inputs not related to sleep occurrences have output probabilities close to 0.

[0138] Given new input, the constructed model outputs a probability of sleep occurring that varies from 0 to 1. The model is updated over time based on new user input and / or feedback data, as well as sensor data related to sleep.

[0139] Other MLMs that can be used for this purpose may include Naive Bayes, Probabilistic Decision Tree, and Probabilistic Support Vector Machine classifiers. Other structures of neural networks may also be incorporated, such as recurrent neural networks, radial basis neural networks, etc.

[0140] 8A-8C illustrate different examples of predictive data models, according to various embodiments. The different examples may include different implementations of the predictive data model as described above with reference to FIG.

[0141] 8A , different multivariate time-series measurements from sensors (e.g., instruments) may be used to predict sleep quality, which may be subjective and / or objective, for a current sleep session in progress. Input data 893-1, 893-2, 893-3, 893-4 to sleep quality prediction data model 894 may include current user sensor data 893-1, previous user sensor data 893-2, other user data 893-3, and heuristics and rules 893-4. Current user sensor data 893-1 may include sensor measurements obtained from the user in the current sleep session in progress and may reflect the user's psychophysiological state (e.g., autonomic function using HRV measurements) and environmental conditions (e.g., outside air temperature). Previous user sensor data 893-2 may include sensor measurements obtained from the user in a previous sleep session and / or objective and / or user self-reports of sleep quality from the previous sleep session. Other user data 893-3 may include sensor measurements taken from multiple other users in prior sleep sessions, along with objective and / or user self-reports of sleep quality from prior sleep sessions. Heuristics and rules 893-4 may include background information and knowledge that may be coded as "if X, then Y" rules.

[0142] The input data 893-1, 893-2, 893-3, and 893-4 are provided to a sleep quality prediction data model 894, which provides an output 895. The output 895 may include a user-self-reported prediction of sleep quality for the current sleep session up to the current time point and / or a predicted objective sleep quality value for the current sleep session up to the current time point. The objective sleep quality value may be composed of several objectively derived values, such as an objectively derived value and / or a weighted combination. In some embodiments, the heuristics and rules 893-4 can provide a starting point and are updated using the other input data 893-1, 893-2, and 893-3 to obtain an accurate and stable prediction as quickly as possible for the current sleep session. The output 895 may be used to determine whether to adjust a sleep intervention strategy to improve sleep quality.

[0143] As shown in FIG. 8B , different sleep intervention strategies can be used to reduce pre-sleep stress and / or improve sleep quality. Input data 896-1, 896-2, 896-3, 896-4 to intervention strategy prediction data model 894 may include user data 896-1, user constraints and preferences 896-2, other user data 896-3, and heuristics and rules 896-4. User data 896-1 may include sensor measurements obtained from the user in a previous sleep session, previous sleep intervention strategies, and / or objective and / or user self-reports of sleep quality from the previous sleep session. User constraints and preferences 896-2 may include user constraints and preferences for intervention actions specific to a particular user. Other user data 896-3 may include sensor measurements obtained from multiple other users during a previous sleep session, along with objective previous sleep intervention strategies used for the user and / or user self-reports of sleep quality from the previous sleep session. Heuristics and rules 896-4 may include background information and rules for intervention actions to maximize sleep quality.

[0144] Input data 896-1, 896-2, 896-3, 896-4 are provided to an intervention strategy predictive data model 894, which provides an output including, in example, a plurality of intervention actions 898-1, 898-2, 898-3, 898-4 forming a sleep intervention strategy. The output may include a recommended course of intervention actions to apply to future sleep sessions forming the sleep intervention strategy. The interventions may be performed in parallel, simultaneously, and / or of variable duration.

[0145] As shown in FIG. 8C , a user can select a specific intervention action within a given sleep intervention strategy and may be provided with an explanation for why the intervention action was selected. Input data 801-1, 801-2, and 801-3 may include a predefined explanation template 801-1, other data 801-2, and a user query 801-3. The predefined explanation template 801-1 may include a template of language and graphics suitable for use as the basis for the explanation. The other data 801-2 may include records of the user's session data (e.g., sensor data, strategies, and self-reports), records of other users' session data, and heuristics and rules encompassing correlations between interventions and self-reports. The user query 801-3 may include a user-selected subset of intervention actions within the selected sleep intervention strategy, which may be associated with past or future sleep sessions.

[0146] The input data of other data 801-2 and user query 801-3 is provided to a key factor prediction data model 803 which outputs key factors to use to select a subset of intervention actions, which are input to a template selection data model 805. The template selection data model 805 pairs the input key factors with templates from predefined explanation templates 801-1, inputs factors from the analysis into the templates, and outputs the pairing of the templates with template population data model 807 which outputs a language-based explanation (e.g., "select X because it gave a boost to Y for similar users") 809, and may also output visualizations and / or plots.

[0147] The above-described system and computer-readable instructions can be used to improve a user's sleep by tracking various factors of the user's sleep and generating a predictive data model that is dynamically updated over time. Based on the dynamic predictive model, the system can be used to predict the occurrence of sleep transitions and increase the probability by developing and implementing sleep invention strategies.

[0148] Embodiments according to the present disclosure include systems, devices, and methods that involve managing the sleep of one or more users. Particular embodiments are directed to acute psychophysiological state manipulation and sleep management systems, and specific implementations thereof.

[0149] FIG. 9 illustrates an example of the effect of intervention on heart rate, HRV over time, and use of an intervention planner to generate a sleep intervention strategy for any particular user, according to various embodiments.

[0150] More specifically, Figure 9 illustrates the acute effect of virtual reality respiratory biofeedback on physiological arousal during a daytime relaxation test, using data from a 49-year-old woman with a sleep disorder during a virtual reality relaxation test. The woman was seated in a reclined position. After initiating respiratory biofeedback (approximately 200 seconds after the start of the test), the subject reduced her breathing rate by approximately 0.1 Hz (6 breaths per minute), as shown in graph 897, improving HRV as indicated by the interbeat intervals shown in graph 8975, and decreasing HR as shown in graph 893. This data demonstrates that behavioral intervention (in this case, virtual reality respiratory biofeedback) can acutely (within seconds) modulate a user's physiological state during wakefulness.

[0151] FIG. 10 illustrates interbeat interval times for a person falling asleep with and without intervention (in this case, virtual reality biofeedback), according to various embodiments. For example, FIG. 10 uses data from a 53-year-old woman with severe insomnia to demonstrate physiological deactivation (here, reflected by enhanced interbeat intervals, e.g., a decrease in heart rate) during virtual reality breathing biofeedback across the transition from wakefulness to sleep. After a night of adaptation, the subject spent two nights in the laboratory, one of which received virtual reality biofeedback intervention to promote sleep. Compared to the control night, virtual reality breathing biofeedback increased pre-sleep heart beat intervals (decreased HR), and the subject approached sleep in a state of physiological relaxation. This data demonstrates that behavioral intervention (in this case, virtual reality breathing biofeedback) across the transition from wakefulness to sleep can enhance relaxation and promote sleep.

[0152] 11A and 11B illustrate the group-level relationship between perceived pre-sleep cognitive arousal and the subsequent time spent falling asleep for individuals with and without insomnia, according to various embodiments. For example, FIGS. 11A and 11B illustrate an example of the relationship between self-reported pre-sleep cognitive arousal and the objective time individuals spend falling asleep, using data from 25 women aged 43-57 with clinical insomnia (shown in FIG. 11B) and 17 women without clinical insomnia (shown in FIG. 11A). Cognitive arousal at bedtime was measured using a standard questionnaire following a stress anticipation procedure. Specifically, the women were told that the next day they would have to give a speech simulating a job interview in front of a panel that would evaluate their performance. The relationship between stress-induced cognitive arousal and objective time to fall asleep was significantly greater (p<0.001) for both women with and without insomnia (group level). These data suggest the importance of focused interventions aimed at reducing pre-sleep arousal levels to improve sleep.

[0153] FIG. 12 illustrates the relationship between physiological pre-sleep activation state (cortisol level) and nighttime polysomnography sleep efficiency, according to various embodiments. For example, FIG. 12 illustrates an example of the relationship between pre-sleep physiological stress levels and polysomnography sleep quality using data from 18 healthy women aged 45 to 51 without clinical sleep disorders. The graph shows the relationship between salivary cortisol collected just before bedtime and nighttime polysomnography sleep efficiency, indicating that women with higher bedtime cortisol levels had lower sleep efficiency. This data suggests that pre-sleep stress levels affect nighttime sleep quality.

[0154] Different sources of information are used to determine the user's psychophysiological activation level and, most importantly, to determine when this activation exceeds what is considered an "adaptive normal range." This is one aspect of the disclosed system, given that awareness of an individual's state of alertness at bedtime is useful for selecting the correct intervention and / or combination of interventions to reduce alertness to an optimal level.

[0155] The optimal level of pre-sleep activation can be determined based on the relationship between pre-sleep physiological activation levels and objective nighttime sleep quality, which is expected to be negative (the more pre-sleep physiological activation, the lower the sleep quality). Additionally, several other behavioral outcomes can be considered, as well as the user's perception of activation.

[0156] In this example, objective sleep efficiency can be used to determine the optimal pre-sleep physiological activation level a user requires to achieve a certain sleep quality, which can be determined based on the relationship between an individual's pre-sleep HRV state and subsequent sleep quality over time.

[0157] 13 is a diagram illustrating the effect of pre-sleep states on sleep, according to various embodiments. In this case, a function was extracted that represents the relationship between pre-sleep HRV states and sleep quality, which can be used to determine a threshold that distinguishes a user's normal pre-sleep hyperarousal state from a reduced, more optimal pre-sleep arousal state. This function can be determined on an individual and / or population-based basis. While a simple hypothetical relationship is shown here, this relationship may change over time and be influenced by multiple factors.

[0158] Resting HRV can be obtained by processing the beat-to-beat variability of the heart rate (low HRV is an indicator of poor autonomic function). It can be calculated over a period of time (e.g., 5 minutes), such as while lying in bed before sleep.

[0159] Sleep efficiency was calculated by dividing the time spent asleep by the total time spent in bed and multiplying by 100. Table 2 shows an example of a hypothetical user's 10 consecutive nights. [Table 2]

[0160] FIG. 14 illustrates a theoretical plot of the relationship between physiological autonomic activation (e.g., HRV) before sleep and sleep efficiency during the subsequent night, according to various embodiments. For example, FIG. 14 shows an example of 10 consecutive nights from a hypothetical user. In the example shown in FIG. 14, the user is considered to be in a state of hyperarousal with an HRV<200 based on evidence that an HRV<200 is most likely to indicate a poor night's sleep (sleep efficiency<85%). In this case, intervention is performed with the goal of raising HRV to a value greater than 200. The curves (HRV and sleep efficiency before sleep) are updated daily, and the thresholds may change as a result.

[0161] Self-report tools (e.g., VAS, questionnaires, etc.) can also be used. For example, to calculate a fixed threshold for an adaptive normal range of a user's perceptual activation, the threshold can be set based on the individual's perception of an altered psychophysiological state. For example, an individual can rate their level of anxiety on a VAS ranging from 11 to 100 mm (ranging from "very low" to "very high") and then assess whether this activation is considered "beyond what they consider normal." Based on population-averaged responses, an adaptive normal range threshold can be determined and stratified based on the individual's demographics.

[0162] Various embodiments are implemented in accordance with the underlying provisional application entitled "AI SLEEPZZZ" (No. 63 / 045304), filed June 29, 2020, entitled "AI SLEEPZZZ," to which benefit is claimed, the general and specific teachings of which are incorporated herein by reference in their entirety. For example, the embodiments in this specification and / or the provisional application may be combined to various degrees (including entirely). Reference may also be made to the experimental teachings and underlying references provided in the underlying provisional application. The embodiments discussed in the provisional application are not intended in any way to limit the overall technical disclosure or any portion of the claimed disclosure, unless otherwise specified.

[0163] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. For example, as used herein, the singular forms "a," "an," and "the" are intended to include the plural forms unless the context clearly dictates otherwise. As used herein, it is further understood that the terms "comprises" and / or "comprising" specify the presence of stated features, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, elements, components, and / or groups thereof. As used herein, the term "and / or" includes any and all combinations of one or more of the associated listed items and can be abbreviated as " / ."

[0164] While various exemplary embodiments have been described above, numerous optional modifications can be made to the various embodiments without departing from the scope of the invention, as set forth in the claims. For example, although described or illustrated with respect to one embodiment, the features and elements so described or illustrated may be applied to other embodiments. As a further example, the order in which various described method steps are performed may often be changed in alternative embodiments, and one or more method steps may be omitted entirely in other alternative embodiments. Optional features of various device and system embodiments may be included in some embodiments and not in other embodiments. Therefore, the foregoing description has been provided primarily for illustrative purposes and should not be construed as limiting the scope of the invention, as defined by the claims.

[0165] Those skilled in the art will recognize that various terms used herein (including the claims) are given their plain meaning in the art unless otherwise indicated. By way of example, this specification describes and / or illustrates aspects useful for implementing the claimed disclosure through various circuits or circuitry, which may be exemplified as or using terms such as block, module, device, system, unit, controller, and / or other circuit type depictions. Such circuits or circuitry, when used in conjunction with other elements, illustrate how particular embodiments may form or implement structures, steps, functions, operations, activities, etc. For example, in certain embodiments described above, one or more modules that may be implemented in the manner illustrated in the figures are discrete logic circuits or programmable logic circuits configured and arranged to perform those operations / activities. In certain embodiments, such programmable circuits are one or more computer circuits that include memory circuits for storing and accessing programs executed as a set (or sets) of instructions (and / or used as configuration data to define how the programmable circuit executes), and the algorithms or processes described herein used by the programmable circuit to execute the associated steps, functions, operations, activities, etc. Depending on the application, the instructions (and / or configuration data) can be configured to be implemented in logic circuitry, with the instructions (whether characterized in the form of object code, firmware, or software) being stored in and accessible from memory (circuitry).

[0166] The various embodiments described above can be implemented together and / or in other ways. One or more of the items depicted in this disclosure can also be implemented separately or in a more integrated manner, as may be useful according to a particular application, or may be removed and / or rendered inoperative in certain cases. In view of the description herein, those skilled in the art will recognize that many modifications may be made without departing from the spirit and scope of the present disclosure.

Claims

1. a memory circuit for storing different patterns based on data indicating a psychophysiological state when a user transitions from a wakeful state to a sleep state and a prediction data model indicating the probability that the user will transition from a wakeful state to a sleep state for each of the patterns; A processing circuit, using features of the data indicative of the user's current psychophysiological state to detect distinct patterns in the predictive data model that correspond to the features and are indicative of a probability that the user will transition from a wakefulness state to a sleep state at a given time on a given date and time; selecting an intervention action corresponding to a characteristic predicted to increase the probability that the user will transition from a wakefulness state to a sleep state at a certain date and time based on the detected pattern; a processing circuit that communicates a message to a user indicating the intervention action; A system comprising:

2. a processing circuit for detecting the pattern in the data by identifying a feature set from among a plurality of feature sets that corresponds to the feature and using the feature set to select a sub-model of the predictive data model; the predictive data model includes a plurality of sub-models that indicate the probability of the user transitioning to a sleep state in response to different intervention actions; the plurality of sub-models are associated with particular Feature Sets of the plurality of Feature Sets; The system of claim 1.

3. the plurality of sub-models relate to different time frames; each feature in the feature set having a weight associated with a probability that the user will transition to a sleep state; The system of claim 2.

4. The system of claim 1 , wherein the processing circuitry modifies the predictive data model based on feedback data indicating whether the user transitions to a sleep state in response to the intervention.

5. The processing circuitry receiving the feedback data in real time; communicating another message in response to the feedback data and the modified predictive data model indicating the modified intervention action. The system of claim 4.

6. The processing circuitry Receive feedback data, In response to receiving the feedback data, Identifying characteristics of the feedback data; Identifying whether the user will respond to the intervention behavior predicted by the predictive data model to increase the probability based on the identified characteristics; modifying the predictive data model for the user associated with the detected pattern in response to an unexpected response based on the input data; The system of claim 4.

7. the intervention is part of a sleep intervention strategy that includes multiple interventions; The plurality of intervention actions are selected from behavioral intervention actions, cognitive intervention actions, neuromodulatory actions, environmental modifications, sensory actions, and combinations thereof. The system of claim 1.

8. The system of claim 7 , wherein the processing circuitry communicates a message indicating the sleep intervention strategy and including a sequence of the plurality of intervention actions.

9. the processing circuitry communicates a plurality of messages including a message indicating the plurality of intervention actions; each of the plurality of messages is selected from the group consisting of a message to the user instructing the user to take each of the intervention actions, and a message to another device for automatically initiating each of the intervention actions at a specific time in accordance with the sleep intervention strategy; The system of claim 7.

10. further comprising an input circuit for receiving data indicative of the user's current psychophysiological state; the input circuitry includes a wearable physiological sensor that senses a physiological signal of the user and another sensor that senses an air measurement; The system of claim 1.

11. further comprising an input circuit for receiving data indicative of the user's current psychophysiological state; the received data is selected from the group consisting of schedule or calendar data, stress levels, general mood, dietary data, health information, exercise data, sleep data, and combinations thereof; The system of claim 1.

12. When executed, the processing circuitry identifying each feature set of a plurality of feature sets of data indicative of a current psychophysiological state of the user; detecting patterns associated with a predictive data model corresponding to the identified feature set and indicative of a probability that the user will transition from a wakeful state to a sleep state at a given time on a given date and time; communicating to the user a message indicating an intervention action corresponding to the pattern predicted to increase the probability that the user will transition from a wakefulness state to a sleep state at a certain date and time based on the detected pattern and the predictive data model; modifying the prediction data model based on feedback data indicating whether the user transitions to a sleep state in response to the intervention action; comprising instructions executable to: Non-transitory recording media.

13. The instruction: detecting a pattern comprising executable instructions to select an associated sub-model of the predictive data model using the identified feature set; communicating another message in response to the feedback data and the modified predictive data model indicating the modified intervention action. comprising instructions executable to: The non-transitory recording medium of claim 12.

14. 13. The non-transitory storage medium of claim 12, wherein the instructions to modify the predictive data model comprise instructions executable to modify a probability weight associated with the intervention action in response to the feature set identified for the user.

15. 13. The non-transitory storage medium of claim 12, wherein the instructions to modify the predictive data model comprise instructions executable to modify the predictive data model for the user over time based on the feedback data and additionally received feedback data indicative of different sleep intervention strategies and respective multiple feature sets.

16. the intervention is part of a sleep intervention strategy that includes multiple interventions; The instruction: communicating a message indicating the sleep intervention strategy and including a sequence of a plurality of intervention actions; modifying the predictive data model, including modifying the sequence of the plurality of intervention actions and one or more of the plurality of intervention actions; comprising instructions executable to: The non-transitory recording medium of claim 12.

17. an input circuit for receiving data indicative of a user's current psychophysiological state; a memory circuit for storing different patterns based on data indicating the psychophysiological state of the user when the user transitions from a wakeful state to a sleep state and a prediction data model indicating the probability of the user transitioning from the wakeful state to a sleep state for each of the patterns; a processing circuit for using features of the data indicative of the user's current psychophysiological state to detect distinct patterns in the predictive data model that correspond to the features and indicate a probability that the user will transition from a wakeful state to a sleep state at a certain date and time, identifying a sleep intervention strategy based on the detected patterns, the sleep intervention strategy including at least one intervention action corresponding to the features predicted to increase the probability that the user will transition from a wakeful state to a sleep state at a certain date and time, and communicating a message to the user indicating the at least one intervention action; A system comprising:

18. 18. The system of claim 17, wherein the memory circuitry includes instructions that, when executed, cause the predictive data model to be generated based on general population trends and publicly available information, and to modify the predictive data model for a user over time using feedback data indicating the success of different sleep intervention strategies for each feature set.

19. 20. The system of claim 17, wherein the at least one message indicating the at least one intervention action further includes an indication of a sequence and timing of the at least one intervention action.

20. 20. The system of claim 17, wherein the processing circuitry communicates at least one message to another device to automatically initiate at least one of the intervention actions at a specific time in response to the sleep intervention strategy.

Images