Smart Pet Bowl

Abstract

This disclosure includes a smart pet bowl for monitoring the feeding behavior of a pet, the smart pet bowl including a pet bowl for holding pet food or water, or for holding a bowl insert for holding pet food or water. The smart pet bowl may also include a load sensor associated with the pet bowl that is capable of sensing changes in the load of pet food or water contained within the pet bowl. The load sensor has a sensitivity of + / - 50 grams or less, and the load data that can be collected therefrom is at a sample rate of 10 to 150 samples per second with continuous time increments of about 0.01 seconds to 5 seconds. The smart pet bowl may also include a data communicator for communicating load data over a computer network, a processor, and a memory for storing instructions that, when executed by the processor, communicate load data over the computer network.

Description

CROSS-REFERENCE TO RELATED APPLICATIONS

[0001]

[0001] This application claims the benefit and priority of U.S. Provisional Patent Application No. 63 / 580550, filed on September 5, 2023, the entire disclosure of which is incorporated herein by reference. BACKGROUND ART

[0002]

[0002] Pet bowls are used by various pet owners or animal caretakers to provide food or water (or other liquids) to animals. Eating or drinking behavior can provide several clues regarding healthy eating habits, and in some cases, eating or drinking behavior can be a suitable tool for detecting potential animal health problems. For example, several visual indicators related to an animal's eating or drinking behavior can be used to provide information regarding the animal's health, including the onset of physical, behavioral, or mental health problems. Unfortunately, these visually noticeable symptoms may only become visually apparent in the middle to late stages of a disease or health problem, and often do not provide sufficient information for proper intervention. Furthermore, animal caretakers, such as pet owners, often lack knowledge about animal behavior to associate eating or drinking behavior with health problems.

[0003]

[0003] Several efforts have been made to track animal eating and drinking behavior, including the use of cameras, scales, and the like. While these devices can be useful for tracking some basic information, such as the amount of food or water consumed, the time of food or water intake, etc., these devices typically provide insufficient information to assess small changes in eating and / or drinking behavior that can provide clues to an animal's health. BRIEF DESCRIPTION OF THE DRAWINGS

[0004] [Figure 1]

[0004] FIG. 1 schematically illustrates an exemplary system including a smart pet bowl for monitoring a pet's eating behavior in accordance with the present disclosure.

[0005] [Figure 2]

[0005] Figure 2 shows an exemplary smart pet bowl according to the present disclosure, including a bowl support and a plurality of bowl inserts.

[0006] [Figure 3A]

[0006] Figure 3A shows an exemplary smart pet bowl feeder according to the present disclosure. [Figure 3B]

[0006] Figure 3B shows an exemplary smart pet bowl feeder according to the present disclosure.

[0007] [Figure 4]

[0007] Figure 4 is a flowchart illustrating an exemplary data collection and processing system and method for monitoring the feeding behavior of pets according to the present disclosure.

[0008] [Figure 5]

[0008] Figure 5 is a flowchart illustrating an exemplary data collection and processing system and method for monitoring pet drinking behavior according to the present disclosure.

[0009] [Figure 6]

[0009] Figure 6 is a more specific flowchart illustrating an exemplary artificial intelligence or machine learning system and method for monitoring the feeding behavior of pets according to the present disclosure. [Figure 7]

[0009] Figure 7 is a more specific flowchart illustrating an exemplary artificial intelligence or machine learning system and method for monitoring the feeding behavior of pets according to the present disclosure. [Figure 8]

[0009] Figure 8 is a more specific flowchart illustrating an exemplary artificial intelligence or machine learning system and method for monitoring the feeding behavior of pets according to the present disclosure. [Figure 9]

[0009] FIG. 9 is a more specific flowchart illustrating an exemplary artificial intelligence or machine learning system and method for monitoring feeding behavior of a pet, according to the present disclosure.

[0010] [Figure 10]

[0010] FIG. 10 is a graph illustrating an embodiment of identifying feeding motion classification, according to the present disclosure.

[0011] [Figure 11]

[0011] FIG. 11 is a flowchart illustrating an exemplary health insight system and method, according to the present disclosure.

[0012] [Figure 12]

[0012] FIG. 12 is a graph illustrating an embodiment of estimating, identifying, and filtering pre-meal and post-meal segments, according to the present disclosure.

[0013] [Figure 13]

[0013] FIG. 13 is a graph illustrating a meal session with identified pet interactions and / or time increments in accordance with the present disclosure. [Figure 14]

[0013] FIG. 14 is a graph illustrating a meal session with identified pet interactions and / or time increments in accordance with the present disclosure. [Figure 15]

[0013] FIG. 15 is a graph illustrating a meal session with identified pet interactions and / or time increments in accordance with the present disclosure.

[0014] [Figure 16]

[0014] FIG. 16 is a graph illustrating an exemplary weight data signal signature obtained from a pet while lapping water from a smart pet bowl, according to the present disclosure.

[0015] [Figure 17]

[0015] FIG. 17 is a graph showing an exemplary load data signal signature obtained from a dog while licking water from a smart pet bowl according to the present disclosure.

[0016] [Figure 18]

[0016] FIG. 18 is a graph showing exemplary data collected from a dog indicating lapping segments, licking segments, and contact with the bowl according to the present disclosure.

[0017] [Figure 19]

[0017] FIG. 19 is a flowchart illustrating an exemplary system and method for processing data collected during animal interaction with a smart pet bowl using an iterative model with normalization logic according to the present disclosure. DETAILED DESCRIPTION OF EMBODIMENTS FOR CARRYING OUT THE INVENTION

[0018]

[0018] The present disclosure relates to animal health and behavior monitoring, and more specifically to devices, systems, methods, and computer program products for measuring, monitoring, processing, recording, and network transferring various physiological parameters and behavioral parameters of animals.

[0019]

[0019] According to embodiments of the present disclosure, a smart pet bowl for monitoring a pet's feeding behavior may include a pet bowl for holding pet food or water, and a load sensor associated with the pet bowl that is capable of sensing changes in the load of the pet food or water contained in the pet bowl. The load sensor may have a sensitivity of + / - 50 grams or less, and the load data that can be collected therefrom may have a sample rate of 10 to 150 samples per second with continuous time increments of about 0.01 to 5 seconds. The smart pet bowl may also include a data communicator for communicating load data over a computer network, a processor, and a memory for storing instructions that, when executed by the processor, communicate load data over a computer network.

[0020]

[0020] In another embodiment, a system for monitoring a pet's feeding behavior may include a smart pet bowl having a pet bowl for holding pet food or water, and a load sensor associated with the pet bowl that is capable of sensing changes in the load of the pet food or water contained in the pet bowl. The load sensor may have a sensitivity of + / - 50 grams or less, and the load data that can be collected therefrom may have a sample rate of 10 to 150 samples per second in continuous time increments of about 0.01 to 5 seconds. In more detail, the smart pet bowl may include a data communicator that communicates load data over a computer network. The system may further include a processor and a memory for storing instructions, the instructions which, when executed by the processor, include receiving load data from the data communicator and identifying feeding behavior that occurs within one or more time increments based on the pet interacting with the pet bowl or the contents of the pet bowl. In this system, in some embodiments, the processor and memory can be mounted and located in the smart pet bowl, while in other embodiments, the processor and memory can be physically located remotely from the smart pet bowl and communicate with a data communicator via a network. In some embodiments, the memory may further include storing instructions, which, when executed by the processor, notify the pet caretaker of feeding behavior or changes in the pet's feeding behavior. Notifying the caretaker could include, for example, warning the caretaker that the pet may have a health problem.

[0021]

[0021] According to embodiments of the present disclosure relating to a smart pet bowl and / or system for monitoring pet feeding behavior, various details relating to the smart pet bowl and / or system may be similar, regardless of whether the processor and / or memory is mounted on the smart pet bowl or located remotely relative to the smart pet bowl. For example, the sample rate, continuous time increments, or both may be controlled onboard by the smart pet bowl and / or by a client device via a computer network. In some embodiments, the smart pet bowl may be capable of filtering out human load data, false triggers, or accidental interactions with the smart pet bowl or its contents. In some embodiments, the smart pet bowl may be able to distinguish between multiple pets in a household with multiple pets. In more detail, the sensitivity, sample rate, and continuous time increments may be established at a level sufficient to identify count-based feeding behaviors, such as wrapping, licking, or biting actions. In other embodiments, the sensitivity, sample rate, and continuous time increments may be established at a level sufficient to allow counting of individual micro-events of feeding behavior within a single time increment or over a continuous period spanning multiple time increments. Individual microevents may include, for example, individual wrapping, individual licking, or individual biting actions. More specifically, sensitivity, sample rate, and successive time increments can be established to a level sufficient to identify duration-based feeding behaviors. Exemplary duration-based feeding behaviors include the pet touching the bowl, moving the bowl, sniffing the food, pausing, eating, wrapping, licking, or a combination thereof. Sensitivity, sample rate, and successive time increments can also be established to a level sufficient to enable successive mapping of time increments in which duration-based feeding behaviors occur or do not occur.In some embodiments, secondary sensors may be included, such as proximity sensors, cameras, microphones, accelerometers, gyroscopes, inertial measurement unit sensors, radar, or combinations thereof. Dogs and cats are examples of pets to which this technology can be utilized.

[0022]

[0022] Regarding the sensitivity of the load sensor, one exemplary prototype has been found to function efficiently with a signal-to-noise ratio of 4 grams, and therefore, in this particular embodiment, any signal generated above 4 grams provides a valid tool for identifying pet feeding behavior. In other systems, lower signal-to-noise ratios of 0.1 grams or less, 0.25 grams or less, or 0.5 grams or less provide a more sensitive system for identifying pet feeding behavior, particularly behavior that does not generate very large forces against the pet food bowl of this disclosure. More specifically with respect to the sample rate, a range of 10 to 150 samples per second has been found to be effective in identifying pet feeding behavior, while an intermediate range of about 15 to 75 samples per second provides sufficient and accurate information to adequately characterize a variety of feeding behaviors.

[0023]

[0023] In some embodiments, the smart pet bowl may be a dog water bowl equipped with a load sensor having a sensitivity of + / - 4 grams or less. In the dog water bowl, the sample rate may be 15 to 75 samples per second, and the time increment may be at least about 0.4 seconds or at least about 0.5 seconds. In some embodiments, the dog water bowl may have a sample rate of 20 to 75 or 35 to 65 samples per second, and a sensitivity of + / - 0.25 grams or less. In other embodiments, the smart pet bowl may be a dog food bowl equipped with a load sensor having a sensitivity of + / - 4 grams or less. In the dog food bowl, the sample rate may be 15 to 75 samples per second, and the time increment may be at least about 0.3 seconds or at least about 0.333 seconds (or 1 / 3 second). In some embodiments, the dog water bowl may have a sample rate of 20 to 75 or 35 to 65 samples per second, and a sensitivity of + / - 0.25 grams or less.

[0024]

[0024] In other embodiments, the smart pet bowl may be a water bowl for cats equipped with a load sensor having a sensitivity of + / - 2 grams or less. In the water bowl for cats, the sample rate may be 15 to 75 samples per second, and the time increment may be at least about 0.4 seconds or at least about 0.5 seconds. In some embodiments, the water bowl for cats may have a sample rate of 20 to 75 or 35 to 65 samples per second and a sensitivity of + / - 0.25 grams or less. In other embodiments, the smart pet bowl may be a food bowl for cats equipped with a load sensor having a sensitivity of + / - 2 grams or less. In the food bowl for cats, the sample rate may be 15 to 75 samples per second, and the time increment may be at least about 0.3 seconds or at least about 0.333 seconds (or 1 / 3 second). In some embodiments, the water bowl for cats may have a sample rate of 20 to 75 or 35 to 65 samples per second and a sensitivity of + / - 0.25 grams or less.

[0025]

[0025] Additional features and advantages of the Smart Pet Bowl(s) and Systems of this Disclosure for monitoring pet feeding behavior, such as feeding and / or drinking behavior, will be described in the following “Modes for Carrying Out the Invention” and “Brief Description of the Drawings,” and will be evident therefrom. The features and advantages described herein are not exhaustive, and many additional features and advantages will be evident to those skilled in the art by considering the drawings and description. Furthermore, it should be noted that the language used herein has been selected solely for readability and illustrative purposes and does not limit the scope of the subject matter of the invention.

[0026]

[0026] According to embodiments of this specification, a smart pet bowl 110 and system 100 for monitoring the feeding behavior of animals, for example, feeding and / or drinking, is shown as an example in Figure 1. In some embodiments, monitoring the feeding behavior of animals can be used for monitoring the health of animals.

[0027]

[0027] The smart pet bowl 110 may be a single unit having, in particular, a bowl support 112 and one or more load sensors 114A arranged to acquire load data relating to the interaction between an animal and / or a person and the smart pet bowl (or its contents). In the illustrated embodiment, the bowl support is bowl-shaped, and therefore pet food and / or water can be added directly to the bowl-shaped bowl support. However, it is understood that the bowl support may have this shape, or even a different shape suitable for receiving a bowl insert 150 molded to receive pet food and / or water. In this case, the bowl support is bowl-shaped. To show that a separate bowl insert may be included to interact with the load sensors(s) of the bowl support, such a bowl insert is shown by dashed lines. This makes it possible to insert more specialized bowls into the common bowl support, such as those shown in Figure 2 below. In any case, the load sensors(s) can be placed in a position that does not interfere with the natural behavior(s) of the animals. Load sensors can automatically detect various feeding behaviors of animals while they are eating.

[0028]

[0028] More specifically, an enlarged exemplary load sensor 114B is shown with a sample circuit which may be used to collect load data when an animal and / or human interacts with the smart pet bowl. Other load sensors and / or circuits may be used, and / or multiple load sensors may be used. Regardless of which load sensors (one or more), the number of load sensors, the arrangement of the load sensors, etc., are used, the load sensors and associated circuits or logic may be configured to be sensitive and fast enough to provide load data over relatively short time increments (0.1 seconds to 5 seconds) suitable for distinguishing count-based feeding behavior and / or duration-based feeding behavior, and in some embodiments even individual microbehaviors. The smart pet bowl may also be able to filter out human load data, false triggers, or accidental interactions with the smart pet bowl or its contents.

[0029]

[0029] Examples of count-based feeding behaviors include wrapping, licking, and / or biting. These count-based feeding behaviors can be counted using individual microevents associated with feeding behavior within a single time increment or over a continuous period spanning multiple time increments. Accordingly, individual microevents can include individual wrapping, individual licking, or individual biting. Examples of duration-based feeding behaviors include touching the bowl, moving the bowl, touching the food with the nose, pausing, eating, wrapping, and / or licking. These duration-based feeding behaviors can be established to a level sufficient to allow for a continuous mapping of time increments in which duration-based feeding behavior occurs or does not occur. The load sensitivity may be fixed based on the architecture of the load sensor(s) 114B, or it may be modifiable, for example, via selected circuitry. In more detail, the sample rate, continuous time increments, or both may be controlled onboard by the smart pet bowl. In other embodiments, the sample rate, continuous time increment, or both may be controlled by a client device via a computer network.

[0030]

[0030] With respect to a system 100 for monitoring the feeding behavior of a pet (one or more), the smart pet bowl 110 can be used in the context of other computer and / or server systems connected by wireless or wired connections. For example, the system for monitoring the feeding behavior of a pet may include, in addition to the smart pet bowl, a client device 130 and an analysis server system 120, each of which communicates via a network 140. In this embodiment, the smart pet bowl, which may in some embodiments include a bowl support 112 and a bowl insert 150, can be filled with pet food or water (or other liquid). In some embodiments, the analysis server system may be implemented using a single server. In other embodiments, the analysis server system may be implemented using multiple servers. In yet another embodiment, the client device may interact with or utilize the analysis server system, and vice versa.

[0031]

[0031] The client device 130 may include, for example, a desktop computer, a laptop computer, a smartphone, a tablet, and / or any other user interface suitable for communicating with the smart pet bowl. The client device may acquire various data from one or more smart pet bowls 110 as described herein, provide data and insights about one or more animals via one or more software applications, and / or provide data and / or insights to the analysis server system 120. The software applications may provide data about the weight and behavior of animals, track changes in data over time, and / or provide health information about animals, as described herein. In some embodiments, the software applications acquire data from the analysis server system for processing and / or display.

[0032]

[0032] The analysis server system 120 can acquire data from various client devices 130 and / or smart pet bowls 110 as described herein. The analysis server system can provide data and insights about one or more animals and / or transmit the data and / or insights to client devices. These insights may include, but are not limited to, insights about the animal's weight and behavior, changes in data over time, and / or health information about the animal, as described herein. In many embodiments, the analysis server system acquires data from multiple client devices and / or smart pet bowls, identifies a cohort of animals in the acquired data based on one or more characteristics of the animals, and determines insights about the animal cohort. Insights about the animal cohort can be used to provide recommendations for specific animals that have characteristics common to the cohort. In some embodiments, the analysis server system provides a portal (e.g., a website) for a veterinarian to access information about a specific animal.

[0033]

[0033] The smart pet bowl can transmit data to a client device 130 and / or an analysis server system 120 for processing and / or analysis. In some embodiments, the smart pet bowl can communicate directly with a non-networked client device 115 without transmitting data over the network 140. The term “non-networked” client device does not mean that the device is not connected via the cloud or other network, but simply suggests that there is a wireless or wired connection that can connect directly to the smart pet bowl. For example, the smart pet bowl and a non-networked client device can communicate via Bluetooth®. In some embodiments, the smart pet bowl can process load sensor data directly. In other embodiments, the smart pet bowl can use the load sensor data to determine whether the smart pet bowl is balanced, unbalanced, and / or properly calibrated (particularly with multiple load sensors). In some cases, automatic or manual adjustment of one or more adjustable load sensors positioned to interact with the contents of the smart pet bowl can be performed. In other cases, since the load sensor is typically positioned not to be in direct contact with the floor and not to support the weight of the bowl support and / or bowl insert, if included, such balancing and / or calibration may be fixed.

[0034]

[0034] Any of the computing devices shown in Figure 1 (e.g., client device 130, analysis server system 120, smart pet bowl 110, and unnetworked client device 115) may include a single computing device, multiple computing devices, a cluster of computing devices, etc. A computing device may include one or more physical processors communicably coupled to memory devices, input / output devices, etc. As used herein, a processor may also be referred to as a central processing unit (CPU). Client devices may be accessed by the animal owner, a veterinarian, or any other user.

[0035]

[0035] In addition, as used herein, a processor may include one or more devices capable of executing instructions that encode arithmetic, logic, and / or I / O operations. In one exemplary embodiment, the processor may implement a von Neumann architecture model and may include an arithmetic logic unit (ALU), a control unit, and / or a number of registers. In some embodiments, the processor may be a single-core processor capable of executing one instruction at a time (or processing a single pipeline of instructions), and / or a multi-core processor capable of executing multiple instructions simultaneously. In some examples, the processor may be implemented as a single integrated circuit, two or more integrated circuits, and / or a component consisting of a multi-chip module in which individual microprocessor dies are included in a single integrated circuit package and thus share a single socket. As described herein, memory refers to volatile or non-volatile memory devices, such as RAM, ROM, EEPROM, or any other device capable of storing data. Input / output devices may include network devices (e.g., network adapters, or any other components that connect a computer to a computer network), PCI (peripheral component interconnect) devices, storage devices, disk drives, sound or video adapters, photo / video cameras, printer devices, keyboards, displays, etc. In some embodiments, a computing device provides an interface, such as an API or web service, that provides some or all of the data to other computing devices for further processing. Access to the interface may be made open and / or secure using any of a variety of techniques, such as by using a client authorization key, depending on the requirements of the specific application of this disclosure.

[0036]

[0036] The network 140 may include a LAN (Local Area Network), a WAN (Wide Area Network), a telephone network (e.g., a Public Switched Telephone Network (PSTN)), a SIP (Session Initiation Protocol) network, a wireless network, a point-to-point network, a star network, a token ring network, a hub network, a wireless network (including protocols such as EDGE, 3G, 4G LTE, Wi-Fi, 5G, WiMAX), the internet, and the like. Various authorization and authentication techniques, such as username / password, OAuth (Open Authorization), Kerberos, SecureID, and digital certificates, may be used to secure the communication. The network connectivity shown in the exemplary pet feeding behavior monitoring system 100 is exemplary, and it should be understood that any means may be used to establish one or more communication links between computing devices.

[0037]

[0037] A feeding behavior monitoring system can automatically track, for example, feeding frequency, feeding type (food and / or water), feeding behavior based on count-based feeding interactions and / or duration-based feeding interactions, and changes in feeding behavior. For example, by collecting historical information related to the collected load data, it is possible to identify when changes occur that may be associated with health or behavioral conditions that may affect the animal, by directly using (or optionally combining with) various other sensor data and / or other historical data of the animal (e.g., age / life stage, sex, reproductive status, physical condition, weight data, etc.) monitoring of these parameters or other parameters.

[0038]

[0038]

[0039]

[0039] Smart pet bowls and systems for monitoring pet feeding behavior can advantageously provide early indicators of potential health conditions, including but not limited to the physical, behavioral, and mental health of animals. Examples of physical health include, but are not limited to, kidney health, urinary health, metabolic health, and digestive health. More specifically, animal diseases that can correlate with behavioral feeding data obtained from the use of smart pet bowls and related systems include, but are not limited to, diabetes, chronic kidney disease, and hyperthyroidism. Based on these potential health conditions, advance notice can be provided to animal owners and / or veterinarians for further diagnosis and treatment.

[0040]

[0040] More specifically with respect to notifications, information about an animal can be transmitted, which may include notifications relating to indicating the animal's behavior, which can be generated based on classified events and / or historical events for the animal. In some embodiments, notifications may be generated based on events of other animals in the same cohort as the animal. Notifications may indicate that an event has occurred and / or indicate one or more inferences about the animal. For example, an animal's water intake behavior can be tracked over time, and if there is an increase in this behavior over time or across a number of events, since this behavior may be associated with diabetes or advanced renal failure or other kidney disease, a notification can be generated to alert the pet's parent of this change in behavior and prompt the caregiver to take the pet to a veterinarian. In some embodiments, a notification is sent when it is determined that a threshold amount of data and / or events has been met. Notifications may be sent to client devices associated with the animal's owner and / or the animal's veterinarian, as described herein. In many embodiments, the notification provides an indication that asks the user to confirm that the detected event is correct. In this way, notifications can be used to obtain ground truth labels about events that can be used to train and / or retrain one or more machine classifiers. Notifications can also be generated and / or transmitted based on the specific animals that perform the events.

[0041]

[0041] With regard to the operation of the smart pet bowl described herein, various user interfaces can be provided to ensure proper installation, configuration, and use of the system for monitoring feeding behavior. These user interfaces can provide instructions to the user, prompt the user for information, and / or provide insights into the behavior and potential concerns of one or more animals. When setting up the system for monitoring feeding behavior, the initialization and positioning of the smart pet bowl may be valuable in ensuring the accuracy of the collected load data. In some embodiments, the smart pet bowl may function better in a climate-controlled indoor environment without direct sunlight. In other embodiments, the smart pet bowl should be placed at a distance of at least one inch from all walls or other obstacles, as insufficient space may cause the smart pet bowl to stick to obstacles and interfere with data or readings. In addition, the smart pet bowl may be placed at a suitable distance from high-vibration items (such as washing machines and dryers) or high-traffic areas, as vibration may cause false and / or inaccurate readings to occur in the weight sensor. In some embodiments, the smart pet bowl may function better on a smooth, level, hard surface. This is because, although soft or uneven surfaces can affect the accuracy of the load sensor, the load sensor can still function and provide usable data under such conditions. In many embodiments, to improve the introduction of the smart pet bowl into the environment, the smart pet bowl may be gradually acclimated to the animal. For example, the smart pet bowl can be placed in the same room as the litter box for several days to allow the animal to adapt to its presence. Once the animal no longer shows discomfort with the presence of the smart pet bowl, the intensity of the smart pet bowl's settings can be reduced to allow the animal to adapt to subtle sounds and lights that may be present during the smart pet bowl's operation, such as dispenser sounds and LED lights.

[0042]

[0042] In some embodiments, multiple user interfaces may be used to configure a system for monitoring feeding behavior. These user interfaces may include a user interface for initiating the smart pet bowl setup process, a user interface for initiating the network setup process, a user interface for connecting to the smart pet bowl via Bluetooth during the setup process, a user interface for confirming the connection to the smart pet bowl via Bluetooth during the setup process, a user interface for connecting the smart pet bowl to a local area network, a user interface indicating that the smart pet bowl is ready to use, a user interface for physically positioning the smart pet bowl and litter box, and / or a user interface for confirming the completion of the setup process. In some embodiments, if multiple animals will be using the same smart pet bowl, profiles can be generated for multiple animals. These profiles can be used to determine the baseline characteristics of each animal and to track the animals' behavior and characteristics over time.

[0043]

[0043] In some embodiments, a user interface may be used to determine the animal profile. Examples of user interfaces for determining the animal profile include a user interface for a start screen for the animal profile determination process, a user interface for an introduction screen for determining the animal profile, a user interface for entering the animal's name, a user interface for entering the animal's sex, a user interface for entering the animal's reproductive status, a user interface for an introduction screen that describes the process of capturing the animal's current physical condition, a user interface for examining a specific body part of the animal, a user interface for examining the animal's profile, a user interface for examining the animal's hips, and / or a user interface for an end screen for the animal profile determination process.

[0044]

[0044] Every animal is unique and behaves in its own way. A system for monitoring feeding behavior can utilize various machine classifiers without adding collars or tools to track and distinguish multiple animals, although in some embodiments additional secondary sensors may also be used. In some embodiments, the user may be asked for information about a particular event, such as identifying which animal used the smart pet bowl. This information can be used to confirm the identity of the animal associated with a particular event and to retrain the machine classifier to improve the accuracy of future results. For example, regarding an animal's feeding behavior, the system may request confirmation of which animal is associated with an event in order to continuously provide the best available insight(s)(one or more). In some embodiments, once the system has developed a unique profile for a particular animal (e.g., after confirming a threshold number), the frequency of future confirmation requests may decrease.

[0045]

[0045] In some embodiments, a user interface for expert advice notifications is used. This user interface may include a user interface that displays a notification indicating that the pet should be monitored for changes in feeding behavior, a user interface that requests confirmation that the smart pet bowl is properly configured, a user interface that requests additional information about the pet's weight, a user interface that requests additional information about the pet's appearance, a user interface that requests additional information about the pet's excretion, and / or a user interface that provides guidance on contacting a veterinarian if changes in the pet's behavior or condition are causing concern.

[0046]

[0046] The feeding behavior monitoring systems and techniques described herein can offer various advantages over existing systems (however, it should be noted that the systems and methods described herein may, in some cases, be used in conjunction with some of these existing monitoring systems). Existing monitoring systems typically rely on microchips implanted in animals, RFID-enabled collars, and / or visual image recognition to identify individual animals. These systems can be highly invasive (e.g., veterinary intervention to implant microchips in specific parts of the animal), and are prone to failure (e.g., microchips may migrate to other parts of the animal, making them difficult to locate, and RFID collars may wear out, get lost, and / or require frequent battery replacement / recharging).

[0047]

[0047] The pet feeding behavior monitoring system according to this disclosure addresses some of the limitations of other existing systems, particularly when some of these systems interfere with the normal behavior of animals. The feeding behavior monitoring system according to this disclosure can identify and track animals without relying on external identification, such as a microchip or RFID collar. For example, wrapping or licking profiles may be distinctly unique to a first pet compared to a second pet, and the smart pet bowl and / or system can be used to distinguish between multiple pets. For example, in some embodiments, the feeding behavior monitoring system described herein can identify animals and their behavior without relying on image or video information, thereby avoiding the use of cameras or human observers, which may influence the typical behavior of animals. That said, cameras and / or other sensors may be used, in some cases, to benefit the functionality of the smart pet bowl, for example, animal recognition, event segmentation, or behavioral classification.

[0048]

[0048] Referring here to Figure 2, an exemplary smart pet bowl 110 is shown with a bowl support 112 and several different bowl inserts 150A to 150D as options for insertion into the bowl support to interact with a load sensor 114. The load sensor is positioned to communicate load data from interaction with the smart pet bowl, more specifically, from interaction with the contents of the smart pet bowl, e.g., pet food and / or water. Thus, when a bowl insert is used, the bowl insert is configured to contact an area adjacent to the load sensor so that the load sensor picks up loads and load changes occurring within the bowl insert. As described above, the bowl support may be a smart pet bowl (without a bowl insert) in some embodiments, or the combination of the bowl support and the bowl insert may be the smart pet bowl. In this case, either is possible as the bowl support is bowl-shaped.

[0049]

[0049] When bowl inserts 150A to 150D are present, any of several configurations can be used, and therefore, in some cases, the bowl support 112 (or base unit) can be made more versatile. For example, different sized bowl supports may exist based on the size and / or type of animal, in which case various bowl inserts may be available together with a more versatile bowl support. Four non-limiting embodiments of the bowl inserts shown in Figure 2 can each interact with a load sensor (and / or other sensors) when that particular bowl insert is placed on the bowl support. Thus, in this embodiment, the bottom surface of the bowl insert is positioned to transmit load during interaction with the feeding pet in order to detect and generate load data for monitoring and / or interpreting health problems. Bowl insert 150A is shown, for example, as a large diameter deep dish that may be suitable for providing a larger amount of pet food and / or water (or other liquid) to the animal. Bowl insert 150B is a shallow dish that reduces the volume for adding pet food and / or water, and may be more suitable for younger pets, such as young adult or young dogs or medium-sized cats. Bowl insert 150C is even shallower with a tapered diameter and may be more suitable for puppies and / or kittens. Bowl insert 150D is a large dish with a maze configuration at the bottom. This configuration may be suitable for slowing down the pet's feeding process for animals that tend to eat too quickly. Regardless of the type of bowl insert used, load data can be collected and analyzed as described herein. In some embodiments, there may be an electrical connection between the bowl support and the bowl insert, thereby providing more information that the smart pet bowl can know what type of bowl is placed on it and, consequently, monitor and / or analyze for health issues. Alternatively, the smart pet bowl may include a place for the user to input the type of bowl insert being used.

[0050]

[0050] More specifically with respect to the smart pet bowl 112, in addition to load sensors (one or more) 114 (and / or other sensors), the smart pet bowl (or bowl support) may include a processor 116 and memory 118. The processor and memory may be capable of controlling the load sensors and receiving load data from the load sensors. The load data may be stored in the memory temporarily or long-term. A data communicator 117 may also be located within the smart pet bowl and capable of communicating load data to another device. For example, the data communicator may be a wireless networking device with a wireless protocol, such as Bluetooth® or Wi-Fi. The data communicator may transmit load data to a physically remote device capable of processing the load data, for example, the analysis server system 120 in Figure 1. The data communicator may also transmit data via a wired connection, using a data port such as a universal serial bus port. Alternatively, the memory slot may accommodate a removable memory card, which may have load data stored thereon and can be physically removed and transferred to another device for uploading or analysis. In one embodiment, the processor and memory may be able to analyze the load data without transmitting the load data to a physically remote device such as an analysis server system.

[0051]

[0051] The smart pet bowl 110 may include a power supply 119. The power supply may be a battery, such as a replaceable battery or a rechargeable battery. The power supply may be a wired power supply that plugs into a wall outlet. The power supply may be a combination of a battery and a wired power supply. The smart pet bowl may be constructed without a camera or image capture device and may not utilize secondary sensors, such as proximity sensors, cameras, microphones, accelerometers, gyroscopes, or inertial measurement unit sensors. Similarly, the animal may or may not have wearable sensors, such as an RFID collar. A system for monitoring the pet's feeding behavior may include tracking the animal's movements while using the smart pet bowl based solely on load sensor data, or optionally by combining load sensor data with one or more of these secondary sensors. The collected data may then be processed to identify the feeding behavior of a particular animal and may optionally be linked to the pet's health characteristics. Based on these characteristics and features, various events can be determined. In some embodiments, various machine learning classifiers can be used to determine these events, as will be described in more detail herein. These events may include, but are not limited to, human interactions, false triggers, and accidental interactions with the smart pet bowl.

[0052]

[0052] Figures 3A and 3B show an alternative smart pet bowl 110 that, in addition to housing the same or similar components as those described in Figures 1 and 2, may also include an associated automatic feeder 160. As shown in Figure 3A, the automatic feeder is integrated with a bowl support 112 that may be molded to be a container for receiving pet food or that can accept a separate bowl insert (not shown). However, as shown in Figure 3B, the automatic feeder is shown to be modular. In this embodiment, different bowl supports or smart pet bowls may be joined with the automatic feeder. In both embodiments, the smart pet bowl includes one or more load sensors 114 as described above. In particular, by combining the automatic feeder with the smart pet bowl, the automatic feeder can dispense pet food on a schedule and / or automatically change pet feeding, e.g., time, amount, type, etc., based on data learned from animal interaction with the smart pet bowl.

[0053]

[0053] Referring here to the systems and methods shown and described as examples in Figures 4 to 10, in particular, the identification of human interactions, preprocessing (normalizing weight values, cleaning data, identifying and / or trimming data, e.g., meal interruptions or unreliable data), meal segments, feeding behaviors, meal sessions, features, etc., are described as examples for understanding this disclosure. For example, a “meal session” can be conceptually divided into one or more individual feeding behaviors in terms of feeding or drinking behavior, which are, as previously stated, “count-based” (wrapping, licking, or chewing actions) and / or “duration-based” (touching the bowl, moving the bowl, touching food with the nose, pausing, eating, wrapping, or licking). In particular, wrapping, licking, or chewing actions can be classified, for example, by count and / or duration. Furthermore, based on count and / or duration, “features” can be developed in the weight data that can be used to identify individuals or sets of feeding behaviors. Load data can be analyzed in the time domain and / or frequency domain. Time domain features may include, but are not limited to, the mean, median, standard deviation, range, and autocorrelation. Frequency domain features may include, for example, the median, energy, and power spectral density. More specifically, load data can be analyzed as total load, as individual loads per load sensor, and / or at the feeding behavior level via a separation algorithm that separates the load data into individual or small groups of feeding behavior interactions.

[0054]

[0054] Time-domain features and / or frequency-domain features may be generated as input for a computer network or other system that acts as a machine classifier for classifying feeding behavior within one or more feeding sessions. The machine classifier can be used to analyze the load data to identify and / or label feeding behavior over a period of time, e.g., 3 seconds of wrapping followed by 2 seconds of licking, or, in some embodiments, to identify micro-events within a feeding behavior time frame in the load data, e.g., even a single licking. Based on the labels, feeding behavior (or even individual animals) can be classified or categorized.

[0055]

[0055] Various machine classifiers can be used, including but not limited to decision trees (e.g., random forests), k-nearest neighbors, support vector machines (SVMs), neural networks (NNs), recurrent neural networks (RNNs), convolutional neural networks (CNNs), stochastic neural networks (PNNs), heuristics, regression, and GBMs (gradient-boosting machines). RNNs can further include (but not limited to) fully recurrent networks, Hopfield networks, Boltzmann machines, self-organizing maps, learning vector quantization, simple recurrent networks, echo state networks, long short-term memory networks, bidirectional RNNs, hierarchical RNNs, stochastic neural networks, and / or genetic scale RNNs. In many embodiments, combinations of machine classifiers can be used. Using more specific machine classifiers when available and general machine classifiers when not can further improve the accuracy of predictions.

[0056]

[0056] More specifically, with respect to “meal session (one or more)”, this can be conceptually divided into multiple feeding behaviors (eating and / or drinking) for classification purposes. For example, some feeding behaviors may be count-based feeding behaviors, e.g., wrapping, licking, or chewing; and / or duration-based feeding behaviors, e.g., touching the bowl, moving the bowl, touching the food with the nose, pausing, chewing / eating, wrapping, licking, or a combination thereof. Wrapping, licking, or chewing behaviors in particular can be classified, for example, by counting and / or duration. Using these feeding behaviors as examples, features can be found in the weighted data for each feeding behavior stage to identify some or all of these specific feeding behaviors resulting from either counting or duration. The weighted data can be analyzed in either the time domain or the frequency domain or both. Time domain features may include, but are not limited to, the mean, median, standard deviation, range, and autocorrelation. Frequency domain features may include, for example, the median, energy, and power spectral density. In some embodiments, load data can be converted to both time-domain and frequency-domain data. For example, time-domain data can be converted to frequency-domain data using various techniques such as the Fourier transform. Similarly, frequency-domain data can be converted to time-domain data using various techniques such as the inverse Fourier transform. In some embodiments, time-domain features and / or frequency-domain features can be identified based on specific peaks, valleys, and / or flat spots within the time-domain and / or frequency-domain data, as described herein. Furthermore, time-domain and / or frequency-domain features can be found for a single load sensor, individual load sensors in a group, and / or all load sensors. Thus, features can be found to assist in classifying feeding behavior using a machine, one or more machine classifiers, or a modeling system.

[0057]

[0057] More specifically, additional features may include, but are not limited to, the standard deviation of the load, the length of the flat spot, the crossover count of the mean, the count of intrinsic peaks, the count of clearly distinguishable load values, the ratio of clearly distinguishable load values ​​to event duration, the count of the maximum load change in individual sensors, the proportion of medium load bins, the proportion of high load bins, the variability of high load bins, the variance of high load bins, the delay or latency of the autocorrelation function, curvature, linearity, the count of peaks, energy, minimum power, power standard deviation, maximum power, maximum variance shift, maximum Kullback-Leibler divergence, Kullback-Leibler divergence time, spectral density entropy, the derivative of the autocorrelation function, and / or the variability of the autoregressive model. Thus, feeding behavior can be classified based on its correlation with the classified features. For example, the selected features can be used as input to a machine classifier for classifying feeding behavior, which may be count-based behavior or duration-based behavior as described above. Feeding behavior may include a label indicating the type of behavior, and / or a confidence metric indicating the likelihood that the label is correct. Unreliable or untrustworthy data can be ignored or removed, for example. Thus, a machine classifier can be trained, for example, with various training data representing animal feeding behavior and ground truth labels with features as input. The system can then be trained using the training data, and the animals can be evaluated using the learning from the training data with the evaluation data. For example, feeding behavior can be classified based on a confidence metric indicating the likelihood that one or more sets of counts or durations were correctly classified. For example, an event can be classified into licking, wrapping, nose-touching the bowl, touching the bowl, pausing, and / or any of the various other feeding behaviors described herein.

[0058]

[0058] Time-domain features and / or frequency-domain features may be generated as input for a computer network or other system that acts as a machine classifier for classifying feeding behavior within one or more feeding sessions. The machine classifier can be used to analyze the load data to identify and / or label feeding behavior over a period of time, e.g., 3 seconds of wrapping followed by 2 seconds of licking, or, in some embodiments, to identify micro-events within a feeding behavior time frame in the load data, e.g., even a single licking. Based on the labels, feeding behavior (or even individual animals) can be classified or categorized.

[0059]

[0059] Various machine classifiers can be used, including but not limited to decision trees (e.g., random forests), k-nearest neighbors, support vector machines (SVMs), neural networks (NNs), recurrent neural networks (RNNs), artificial neural networks (ANNs), convolutional neural networks (CNNs), stochastic neural networks (PNNs), heuristics, regression, and optical gradient boosting machines (GBMs). RNNs can further include (but not limited to) fully recurrent networks, Hopfield networks, Boltzmann machines, self-organizing maps, learning vector quantization, simple recurrent networks, echo state networks, long-term short-term memory networks, bidirectional RNNs, hierarchical RNNs, stochastic neural networks, and / or genetic scale RNNs. In many embodiments, combinations of machine classifiers can be used. Using more specific machine classifiers when available and general machine classifiers when not can further improve the accuracy of predictions.

[0060]

[0060] Referring more specifically to Figure 4, this flowchart shows an exemplary data collection and processing method 200 and associated systems that may be implemented when monitoring the feeding behavior of a pet. For example, when monitoring feeding behavior, the steps may include behavior classification modeling 230, meal segment identification 240, and / or feeding repetition modeling 250, which are performed in any order. For example, behavior classification modeling may be in the form of a build or score activity classification model and may include aggregating data using rolling time windows of, for example, about 0.01 seconds to about 5 seconds, about 0.05 seconds to about 3 seconds, or about 0.1 seconds to about 1 second. Processing may also include meal segment identification. Meal segment identification may include, for example, identifying the start and end points of a meal session, calculating the food provided and the remaining food after the meal session, and / or trimming non-meal segments for the meal session. In some embodiments, processing may also include feeding repetition modeling, which may be based on a build or score of a feeding (or drinking) repetition regression model. Repetitive eating modeling may involve providing aggregated data, for example, at the meal session level.

[0061]

[0061] When modeling pet feeding behavior, several modeling considerations can be made to determine the type of load data signature that may be useful in characterizing the feeding behavior of a particular type of animal, such as a cat or a dog. For example, when modeling animal feeding behavior, a number of different individuals of the same species are typically used. "Truth data" may be collected for comparison purposes so that it can be correlated with load sensor data collected for individual feeding behaviors as well as for the overall modeling of the smart pet bowl of this disclosure. Truth data can be collected in various ways, such as real-time observation, but video recording works well because it allows technicians to pause, slow down, rewind, etc., when carefully examining feeding behavior. In some cases, truth data may include a combination of video recording and sensor data, provided that the sensor data is found to be reliable. Thus, truth data can be compared to "training data" collected simultaneously from multiple animals of the same species, which may be based on data collected using load sensors (one or more). Therefore, correlation can be performed based on load sensor signals of pet feeding signature behaviors that are consistent with observable pet feeding behavior data collected by video recording. Ground truth data labels (from video) and training data (from load sensors and unedited (raw) video) can be correlated to "train" the smart pet bowl so that a model can be built. In other words, training can be performed to determine the correct load sensor data, and unedited video can be correlated with specific feeding behaviors. For example, using ground truth data from video and correlated training data, load sensor signatures can be determined for feeding behaviors such as eating, wrapping, licking, and dropping food. Once those feeding behaviors are determined for identifiable load sensor signals specific to those behaviors, "test data" can be collected again using groups of the same type of animals, e.g., dogs, cats, etc., to test various models that return results that accurately characterize the feeding behaviors.Depending on the extent to which the ground truth data is inconsistent or different from the data collected using load sensors, that data can be cleaned up from the entire dataset, for example, by being removed from the dataset. Again, when determining the appropriate model, the video "ground truth data" can be used again to verify that the model returns good feeding behavior results that are accurate enough to be useful. As one embodiment, any of the numerous models described elsewhere in this specification can be used.

[0062]

[0062] In some more detailed embodiments, the data collection and data processing 200 when monitoring pet feeding behavior may include performing several additional steps that are typically done before monitoring feeding behavior. For example, human interactions 210 with the smart pet bowl that may appear in the data as load signals unrelated to animal feeding, as described above, may be identified and removed from the feeding segment before processing according to 230, 240, and / or 250. In more detail, load signal data may be preprocessed to clean up the data in order to further improve the reliability of the feeding segment data (220). For example, "ground truth data" collected from video recordings may be compared to identify times when there is human interaction with the smart pet bowl. When these types of load signals are detected, they may be removed from the dataset as being related to human interactions, e.g., placing the bowl down, lifting the bowl, pouring food into the bowl, etc. These types of interactions typically have very different frequencies and amplitudes from those perceived during normal pet feeding behavior. Preprocessing may include normalizing load values ​​and / or identifying and cleaning data that is not particularly relevant to feeding behavior from the dataset to be analyzed, such as identifying and trimming irrelevant data collected, identifying interruptions in feeding sessions, and / or removing data that may be less reliable than any data collected. Again, when training and / or testing the smart pet bowl for data preprocessing, the “ground truth data” can be compared to the training and / or test data to identify usable load sensor frequency responses and load sensor signals corresponding to less reliable load sensor frequency responses.

[0063]

[0063] Modeling may be used based on machine learning or artificial intelligence. One example of how data can be cleaned to provide a more representative sample for learning about pet feeding behavior is described in more detail in Example 1 below. Identification of human interactions may include, for example, identifying human or other anomalous interactions with the smart pet bowl or its contents before meals, identifying human or other anomalous interactions with the smart pet bowl or its contents after meals, and / or trimming these human or other anomalous interactions from the collected load data. Further details regarding identifying and trimming human interactions are described in more detail in relation to Example 1 below.

[0064]

[0064] It should be noted that while the systems and methods described in Figure 4 above and below use other flowcharts, many other methods may be used to perform the operations associated with these methods. For example, the order of some blocks may be changed, some blocks may be combined with other blocks, one or more blocks may be repeated, and / or some of the described blocks may be optional. The methods may be performed by processing logic that may include hardware (circuits, dedicated logic, etc.), software, or a combination of both. The methods or processes may be implemented and performed as instructions on a machine, and the instructions may be contained on at least one computer-readable medium or one non-temporary machine-readable storage medium.

[0065]

[0065] Referring here to Figure 5, a more detailed flowchart of data collection and processing 300 for counting drinking repetitions, or "reps," is shown as an example. In particular, there are other ways of organizing the counting of drinking repetitions, but this embodiment provides one acceptable method that may be carried out. According to this embodiment, raw data 310 is collected and some data preprocessing 320 can be performed, such as identifying human interactions for removal, as described according to Figure 4 in some embodiments. Preprocessing may also include identifying interruptions based on the calculation of normalized weight values ​​with modeling or the use of a machine classifier. Machine classifiers may include, but are not limited to, decision trees (e.g., random forests), k-nearest neighbors, support vector machines (SVMs), neural networks (NNs), recurrent neural networks (RNNs), artificial neural networks (ANNs), convolutional neural networks (CNNs), stochastic neural networks (PNNs), heuristics, regression, optical gradient boosting machines (GBMs), etc. For example, heuristic modeling can be used, which may include any form of computer-based problem-solving or discovery that produces an imperfect but sufficient result as an approximation when pursuing or searching for a solution to a problem. Heuristic methods may allow the discovery process to be accelerated to reach a satisfactory result. Thus, by removing identified interruptions, even fast approximations can typically improve the data collected for subsequent processing. In this embodiment, after processing (or a part thereof) is complete, the data processing may include behavioral classification modeling 330 and wrapping duration and licking duration 350 as separate components, and / or, in other embodiments, the data processing may include iterative modeling 340 and a combination of wrapping and licking iterations 360. This processed data can then be subjected to normalization logic 370 to yield data related to wrapping iterations 380 and licking iterations 390.

[0066]

[0066] More specifically with respect to modeling, for example, behavioral classification modeling and / or repetition modeling described by the embodiments above, classifying feeding and / or drinking behaviors can also be established by other modeling protocols. Regardless of the type of modeling used, the smart pet bowl can collect data based on different load signatures captured and provide information about what type of behavior is occurring at a given time during a meal session. This can lead to monitoring of feeding behavior and, in some cases, the generation of insights related to the pet's health and / or well-being.

[0067]

[0067] Exemplary feeding behaviors that can be captured may be grouped into several larger categories, e.g., animal interactions, repetitions (reps), ingestion, and / or interruptions (one or multiple). Table 1 shows some exemplary, non-limiting behaviors that can be captured using the collection of load data from the smart pet bowl during a feeding session, as follows: [Table 1]

[0068]

[0068] Figures 6 to 9 show several exemplary artificial intelligence (AI) processing protocols that can be used for multiple types of animals regarding feeding and / or drinking. In these embodiments, onboard data entered by the animal caretaker or pet owner can be combined with more continuous data collected by a smart pet bowl or system for monitoring the pet's feeding behavior. By definition, Figures 6 to 9 show various types of data that may be collected and processed by a computer or computer network. Other types of data collection and / or processing locations may be implemented in ways other than those shown in these embodiments. For example, the types of data collected as shown in Figures 6 to 9 may include data relating to an animal's feeding session (denoted as "s" in the figures), the animal's interaction with the smart pet bowl (denoted as "i"), time increments or time windows (denoted as "t"), or a combination of interaction / time increments (denoted as "i / t"). More specifically, the calculation or processing of data may be performed at any location, e.g., a smart pet bowl, a local computer, a local client device or smartphone (or tablet), any location suitable for edge computing (closer to the data source than a centralized server or cloud-based location), a centralized server, or a process represented by the cloud. Often, the “cloud” symbol is used as an example to indicate that the processing in this embodiment is cloud-based, but may take place at any location. Various calculation methods for general calculation or processing are also illustrated in Figures 6–9, including cloud computing (denoted “c”) and / or edge computing (denoted “e”).

[0069]

[0069] Referring more specifically to Figure 6, the system 400 (or smart pet bowl) for monitoring the feeding behavior of a pet may be set up by a user onboard 410, which includes a general user setup. The general setup may involve inputting information such as the animal species, animal breed, animal weight, animal physical condition score, animal reproductive status, animal age, and / or other information that may be relevant to the collection and processing of relevant data related to feeding. If the smart bowl is suitable for pet food or water, what is used may also be input by the user, or alternatively, the smart pet bowl may sense the type of food or water (or other liquid) that may be present. If the smart pet bowl is for a food process (430), the device registers this event classification, and in this embodiment, the process may be used to determine whether the load sensor of the smart pet bowl is triggered as a result of a human event or a pet event. If the smart pet bowl is for a water process (440), the device registers this event classification, and in this embodiment, the process may be used to determine whether the load sensor of the smart pet bowl is triggered as a result of a human event or a pet event. More detailed artificial intelligence and other computer processing are described in Figures 8 and 9 below.

[0070]

[0070] As shown in Figure 7, the smart pet bowl can be set up, for example, using an extended cat feeding process 450. In particular, cloud computing (c) and edge computing (e) are shown only as examples, as previously stated. This process may be used instead of the food pet bowl process shown in Figure 6 430. In this extended process (or system), in addition to the load sensors shown and described earlier, secondary sensors can be used for pet detection, for example, by scanning or otherwise capturing images and / or pet faces. The type of food can be selected to fill the bowl, or the bowl can be filled directly without determining the type of food, e.g., one type of food available. In this case as well, human events can be distinguished from pet events, but in this embodiment, the system or method can utilize cat identification (cat ID) to match the face of a specific cat using a captured image as a preliminary step to registering a pet event for a particular cat. In some embodiments, the smart pet bowl of the present disclosure can provide personalized insights with respect to individual animals, e.g., individual pets in a household with multiple pets. Therefore, the smart pet bowl may be equipped with one or more secondary sensors, such as proximity sensors, cameras, microphones, accelerometers, gyroscopes, inertial measurement unit sensors, and / or radar. In other embodiments, pets can be identified using load data during feeding and / or drinking, as multiple animals may have different feeding behavior profiles that may be distinguishable in some embodiments. In some more specific embodiments, the identification of cats or other animals can be done by image recognition (embedding from images) using a camera, and / or the identification of dogs or other animals can be done using load sensor data with modeling that closely approximates real-world data visually collected for verification.

[0071]

[0071] Figures 8 and 9 provide additional details regarding AI and / or other protocols that can be used to process human events and / or pet events. For example, as shown in Figure 8, after the smart pet bowl is set up and the device registers relevant information (510) entered by the user or detected by the smart pet bowl or peripheral devices, such as cat or dog, pet food details, animal weight, animal breed, physical condition score, reproductive status, age, etc., the system 500 (or smart pet bowl) for monitoring the feeding behavior of the pet can separate human events 520 from pet events 530 for processing. Human events detailed in this embodiment may include filling the smart pet bowl, washing the smart pet bowl, moving the smart pet bowl, lifting, etc. Pet events are more fundamental to the invention of this disclosure, as human events are primarily identified to trim or remove from the collected data in order to more accurately understand the feeding behavior of the animal. Pet events that may be relevant to a dog eating pet food may be classified using load data collected and analyzed as described herein. More specifically, exemplary feeding behavior classifications may include feeding, wrapping, licking, touching food with the nose (moving food with the nose), dropping food, touching the bowl, and / or pausing eating. For a feeding cat, feeding behavior classifications may include, for example, feeding, wrapping, licking, touching food with the nose (moving food with the nose), dropping food, gobbling, touching the bowl or touching the food, and / or pausing eating. Feeding may be further characterized, for example, by food intake, feeding rate, interaction with food, and / or duration of eating. When collecting this type of data, feeding behaviors may be processed, for example, based on behavior repetition counts and / or event duration.

[0072]

[0072] As shown in Figure 9, here again, after the smart pet bowl is set up and the device registers relevant information (610), such as cat or dog, pet food details, animal weight, animal breed, etc., entered by the user or detected by the smart pet bowl or peripheral devices, the system 600 (or smart pet bowl) for monitoring the pet's feeding behavior can separate human events 620 from pet events 630 for processing. Human events detailed in this embodiment may include filling the smart pet bowl, washing the smart pet bowl, moving the smart pet bowl, lifting it, etc. Pet events are more fundamental to the invention of this disclosure, on the other hand, as human events are mainly identified to trim or remove from the collected data in order to more accurately understand the animal's feeding behavior. Pet events that may relate to a dog drinking pet food may be classified using load data collected and analyzed as described herein. More specifically, exemplary drinking behavior classifications may include wrapping, licking, pausing to drink, and / or any other behaviors associated with drinking. For cats, drinking behavior classifications may also include, for example, wrapping, licking, pausing to drink, and / or other identified behaviors that may be associated with drinking in cats. When collecting this type of data, drinking behaviors may be processed, for example, based on behavior repetition counts and / or event duration.

[0073]

[0073] Figure 10 shows an example of feeding behavior classification, which may include identifying actions related to any of the following activities: dropping food, touching food with the nose, taking out contents (feeding), touching the bowl, and / or other activities related to feeding. More specifically, Figure 10 includes data collected and classified (based on load sensor data) on how dogs interact with the smart pet bowl using continuous time increments of 0.3 seconds, e.g., a rolling time window of 0.3 seconds. In other words, every 0.3 seconds, one or more actions are identified and characterized. For the sake of simplicity in describing this example, three of the possible actions are identified, including pausing feeding (p), taking out contents (r), and touching the bowl (t). This dataset does not show any exclusion of human interaction, which may be indicated by the large load fluctuations shown on the left side of the graph.

[0074]

[0074] Figure 11 shows an exemplary system or method 700 that can be used to generate health insights that may be obtainable when collecting feeding data using the smart pet bowl or related system of the present disclosure. In this embodiment, various logic systems or ensemble logics can be combined when processing data associated with feeding events 750 and data collected over various time ranges 760. For example, feeding events can be registered based on unedited classification. These feeding events can be registered using, for example, artificial intelligence and / or machine learning (AI / ML) 710. Machine classifiers or other systems can be used to model feeding, drinking, repetition counts, duration, etc. Exemplary machine classifiers include decision trees (e.g., random forests), k-nearest neighbors, support vector machines (SVMs), neural networks (NNs), recurrent neural networks (RNNs), convolutional neural networks (CNNs), stochastic neural networks (PNNs), heuristics, regression, and optical gradient boosting machines (GBMs).

[0075]

[0075] The normalization logic 720 can process unedited classifications to generate predictions. In some embodiments, user input and / or labels can be taken into consideration and, where appropriate, the normalization logic predictions may be overridden. With respect to data collected over various time ranges, the system or method can utilize aggregation logic 730. The aggregation logic matches the normalization predictions to generate aggregated results. In some embodiments, the aggregated results can then be used in conjunction with a specific decision tree logic 740. Such decision tree logic can alert animal managers or pet owners to relevant information or insights based on their feeding behavior, for example. In one embodiment, the decision tree logic may be used as part of a decision tree logic system, which can be used with other logic systems to transform data into meaningful insights to give pets a voice about their health and well-being, and to enable pet owners to feel confident in decisions made on behalf of their pets.

[0076]

[0076] Examples of information that may be obtainable from data collected using the smart pet bowl and / or system of the present disclosure include event information or insights, daily information or insights, weekly information or insights, monthly information or insights, and yearly information or insights. For example, event information that can be given to pet owners may include the time spent on each action, the number of repetitions of each action, the number of food or water interactions, the grams ingested per chewing action, and the total grams ingested. Examples of daily information may include the time of feeding, the total intake for a given day, the average duration of feeding, the average rate of feeding, and the average number of events. Monthly information may include feeding trends, total intake, the average duration of feeding, the average rate of feeding, and the average number of events per day.

[0077]

[0077] Health and behavioral insights that may be associated with changes from this type of data include dietary or flavor preferences, stress and anxiety levels, acute or chronic illnesses, dental problems, etc. These types of information and insights collected, assessed, and reported can give pet owners confirmation that they are managing their pet's characteristics. As an example of how this may work in some cases, pets create a feeding signature using a bowl. Changes in the feeding signature can be detected and notified to the pet owner. With notification, the pet owner can make adjustments in the best interests of the pet.

[0078]

[0078] It will be understood that all of the methods and procedures disclosed herein can be implemented using one or more computer programs, components, and / or program modules. These components may be provided as a series of computer instructions on any conventional computer-readable or machine-readable medium, including volatile or non-volatile memory, e.g., RAM, ROM, flash memory, magnetic or optical disks, optical memory, or other storage media. The instructions may be provided as software or firmware and / or implemented in whole or in part in hardware components, e.g., ASICs, FPGAs, DSPs, or any other similar devices. The instructions may be configured so that when a series of computer instructions is executed by one or more processors, it performs or facilitates the performance of all or part of the methods and procedures disclosed. As will be understood by those skilled in the art, the functions of the program modules may be combined or distributed as necessary in the various embodiments of this disclosure.

[0079]

[0079] In accordance with the disclosures herein, the following embodiments illustrate some embodiments of the Art. [Example 1] An exemplary smart pet bowl for monitoring pet feeding behavior, i) a pet bowl for holding pet food or water, or ii) a pet bowl insert for holding pet food or water, A load sensor associated with the pet bowl, capable of sensing changes in the load of pet food or water contained within the pet bowl, having a sensitivity of + / - 50 grams or less, and the load data that can be collected therefrom having a sample rate of 10 to 150 samples per second in continuous time increments of approximately 0.01 seconds to 5 seconds, A data communication device that communicates the load data via a computer network, Processor and A memory that stores instructions for communicating the load data via the computer network when executed by the processor, A smart pet bowl equipped with [features]. [Example 2] The smart pet bowl according to Embodiment 1, comprising a bowl support for housing the load sensor and a bowl insert for housing the pet food or water, and configured to transmit a load from the bowl insert to the load sensor. [Example 3] The smart pet bowl according to Embodiment 1 or 2, wherein the smart pet bowl is a single, integrated pet bowl configured to directly contain pet food or water, and the single, integrated pet bowl also houses a load sensor at a position where the animal's interaction with the pet food generates load data. [Example 4] The smart pet bowl according to any one of Examples 1 to 3, wherein the sample rate, continuous time increment, or both are controlled onboard by the smart pet bowl. [Example 5] The smart pet bowl according to any one of Examples 1 to 4, wherein the sample rate, continuous time increment, or both are controlled by a client device via the computer network. [Example 6] The smart pet bowl according to any one of Examples 1 to 5, wherein the smart pet bowl is capable of excluding human load data, false triggers, or accidental interactions with the smart pet bowl or its contents. [Example 7] The smart pet bowl according to any one of Examples 1 to 6, wherein the smart pet bowl can distinguish between multiple pets in a household with multiple pets. "Example 8" The smart pet bowl according to any one of Examples 1 to 7, wherein the sensitivity, sample rate, and continuous time increment are established at a level sufficient to identify count-based feeding behavior selected from wrapping, licking, or biting. [Example 9] The Smart Pet Bowl according to Example 8, wherein the sensitivity, sample rate, and continuous time increments are established at a level sufficient to allow for the counting of individual microevents of feeding behavior within a single time increment or over a continuous period spanning multiple time increments, wherein the individual microevents include individual wrapping, individual licking, or individual biting actions. [Example 10] The smart pet bowl according to any one of Examples 1 to 9, wherein the sensitivity, sample rate, and continuous time increment are established to a level sufficient to identify duration-based feeding behaviors selected from touching the bowl, moving the bowl, touching the pet food with the nose, pausing, feeding, wrapping, licking, or a combination thereof. [Example 11] The smart pet bowl according to Example 10, wherein the sensitivity, sample rate, and continuous time increments are established at a level sufficient to enable continuous mapping of the time increments during which the duration-based feeding behavior occurs or does not occur. [Example 12] A smart pet bowl according to any one of Examples 1 to 11, further comprising a secondary sensor selected from a proximity sensor, camera, microphone, accelerometer, gyroscope, inertial measurement unit sensor, radar, or a combination thereof. [Example 13] The smart pet bowl according to any one of Examples 1 to 11, wherein the smart pet bowl is a water bowl for dogs, the load sensor has a sensitivity of + / - 4 grams or less, the sample rate is 15 to 75 samples per second, and the time increment is at least about 0.4 seconds. [Example 14] The smart pet bowl according to any one of Examples 1 to 11, wherein the smart pet bowl is a dog food bowl, the load sensor has a sensitivity of + / - 4 grams or less, the sample rate is 15 to 75 samples per second, and the time increment is at least about 0.3 seconds. [Example 15] The smart pet bowl according to any one of Examples 1 to 11, wherein the smart pet bowl is a water bowl for cats, the load sensor has a sensitivity of + / - 2 grams or less, the sample rate is 15 to 75 samples per second, and the time increment is at least about 0.4 seconds. "Example 16" The smart pet bowl according to any one of Examples 1 to 11, wherein the smart pet bowl is a cat food bowl, the load sensor has a sensitivity of + / - 2 grams or less, the sample rate is 15 to 75 samples per second, and the time increment is at least about 0.3 seconds. [Example 17] An exemplary system for monitoring the feeding behavior of pets, Smart Pet Bowl and Processor and Memory for storing instructions, Equipped with, The aforementioned smart pet bowl is i) a pet bowl for holding pet food or water, or ii) a pet bowl insert for holding pet food or water, A load sensor, which is associated with the pet bowl and capable of sensing changes in the load of pet food or water contained in the pet bowl, has a sensitivity of + / - 50 grams or less, and the load data that can be collected from the load sensor is collected at a sampling rate of 10 to 150 samples per second in continuous time increments of approximately 0.01 seconds to 5 seconds, A data communication device that communicates the load data via a computer network, including, When the instruction is executed by the processor, Receiving the load data from the data communication device, Identifying feeding behaviors that occur within one or more of the time increments based on the pet's interaction with the pet bowl or its contents, A system that includes this. [Example 18] The system according to 17, wherein the processor and the memory are mounted and arranged in a smart pet bowl. [Example 19] The system according to Embodiment 17 or 18, wherein the processor and the memory are physically located remotely from the smart pet bowl and communicate with the data communication device via a network. [Example 20] The system according to any one of Examples 17 to 19, wherein the memory stores instructions, and when the instructions are executed by the processor, the instructions further include excluding load data of interactions with the smart pet bowl or its contents if it is determined that the interaction is due to human interaction, a false trigger, or an accidental interaction. [Example 21] The system according to any one of Examples 17 to 20, wherein the feeding behavior is a count-based feeding behavior selected from wrapping, licking, or biting. [Example 22] The system according to any one of Examples 17 to 21, wherein the feeding behavior is a duration-based feeding behavior selected from touching the bowl, moving the bowl, touching the pet food with the nose, pausing, eating, wrapping, licking, or a combination thereof. [Example 23] The system according to any one of Embodiments 17 to 22, wherein the memory stores instructions, and the instructions further include notifying the pet's caretaker of the feeding behavior or a change in the pet's feeding behavior when the processor executes the instructions. [Example 24] The system according to Example 23, wherein notifying the administrator includes warning the administrator that the pet may have a health problem. [Example 25] The system according to any one of Examples 17 to 24, wherein the smart pet bowl further includes secondary sensors, including a proximity sensor, a camera, a microphone, an accelerometer, a gyroscope, an inertial measurement unit sensor, or a combination thereof. [Example 26] The system according to any one of Examples 17 to 25, wherein the smart pet bowl is a water bowl for dogs, the load sensor has a sensitivity of + / - 4 grams or less, the sample rate is 15 to 75 samples per second, and the time increment is at least about 0.4 seconds. [Example 27] The system according to any one of Examples 17 to 25, wherein the smart pet bowl is a dog food bowl, the load sensor has a sensitivity of + / - 4 grams or less, the sample rate is 15 to 75 samples per second, and the time increment is at least about 0.3 seconds. [Example 28] The system according to any one of Examples 17 to 25, wherein the smart pet bowl is a water bowl for cats, the load sensor has a sensitivity of + / - 2 grams or less, the sample rate is 15 to 75 samples per second, and the time increment is at least about 0.4 seconds. [Example 29] The system according to any one of Examples 17 to 25, wherein the smart pet bowl is a cat food bowl, the load sensor has a sensitivity of + / - 2 grams or less, the sample rate is 15 to 75 samples per second, and the time increment is at least about 0.3 seconds. [Example 30] The system according to any one of Examples 17 to 29, wherein the system can distinguish between multiple pets in a household with multiple pets. ●Definition

[0080]

[0080] As used herein, “about,” “approximately,” and “substantially” are understood to refer to a range of numbers, for example, a range of -10% to +10% of the referenced number, a range of -5% to +5% of the referenced number, a range of -1% to +1% of the referenced number, or a range of -0.1% to +0.1% of the referenced number. All numerical ranges herein should be understood to include all integers, whole numbers, or fractions within that range. Furthermore, these numerical ranges should be interpreted as supporting claims that cover any number or subset of a number within that range. For example, a disclosure of 1 to 10 should be interpreted as corresponding to ranges such as 1 to 8, 3 to 7, 1 to 9, 3.6 to 4.6, 3.5 to 9.9, etc.

[0081]

[0081] When used in the present disclosure and the appended claims, the singular forms "a," "an," and "the" also refer to multiple objects unless the context indicates otherwise. Thus, for example, a reference to "a component" or "the component" includes two or more components.

[0082]

[0082] The terms “comprise,” “comprises,” and “comprising” should be interpreted as not being exclusive but potentially encompassing each other. Similarly, the terms “include,” “including,” and “or” should all be interpreted as potentially encompassing each other unless such interpretation is clearly prevented by the context. Accordingly, disclosures of embodiments using the term “comprising” include disclosures of embodiments that “consist essentially of” the specified components, and disclosures of embodiments that “consist of” the specified components.

[0083]

[0083] The term "and / or" used in the context of "X and / or Y" should be interpreted as "X" or "Y" or "X and Y". Similarly, "at least one of X or Y" should be interpreted as "X" or "Y" or "both X and Y".

[0084]

[0084] In use herein, the terms “example” and “such as” are merely illustrative and descriptive, and should not be considered exclusive or comprehensive, especially when they are followed by a list of terms.

[0085]

[0085] The terms “pet” and “animal” are used synonymously herein and mean any animal on which the Smart Pet Bowl or related systems of this disclosure can be used, non-limiting examples of which include cats, dogs, rats, ferrets, hamsters, rabbits, iguanas, pigs, or birds. A pet can be any suitable animal, and this disclosure is not limited to any particular pet animal. The Smart Pet Bowl may or may not be particularly adapted to a particular type of animal, e.g., a dog or a cat, and / or may or may not be particularly adapted to a particular feeding behavior, e.g., liquids (water, milk, etc.) or food (dry kibble, wet food, or semi-wet food, etc.).

[0086]

[0086] As used herein, range is an abbreviation to avoid the need to enumerate and list every single value within that range. Any suitable value within the range can be selected to be appropriate as an upper limit, lower limit, or end of the range, and thus should be interpreted flexibly to include the numerical value explicitly cited as limiting the range, and also to include the individual numerical values ​​or subranges contained within the range, where numerical values ​​and subranges are explicitly cited. For example, the numerical range "about 1% to about 5%" should be interpreted to include the multiple values ​​explicitly stated as about 1% to about 5%, and also to include the individual values ​​and subranges within the indicated range. Thus, this numerical range includes the individual values, e.g., 2, 3.5, and 4, as well as the subranges, e.g., 1 to 3, 2 to 4, and 3 to 5. The same principle applies to ranges that represent a single numerical value. Furthermore, such interpretations should apply regardless of the range of the range or feature described.

[0087]

[0087] The terms “example(s)” or “embodiment(s)” should not be considered exclusive or comprehensive, especially when the terms are subsequently listed.

[0088]

[0088] The methods, storage media, smart pet bowls, and systems disclosed herein are modifiable as will be recognized by those skilled in the art, and are not limited to the specific methods, protocols, reagents, etc. described herein. Furthermore, the terms used herein are for the sole purpose of describing specific embodiments and are not intended to, and do not limit, the scope of disclosure or claims.

[0089]

[0089] Unless otherwise specified, all technical and scientific terms, specialized terms, and acronyms used herein have meanings that are generally understood by those skilled in the art of one or more fields of the present invention or one or more fields in which the terms are used. Any composition, method, product, or other means or materials similar to or equivalent to those described herein may be used in carrying out the present invention, but specific compositions, methods, products, or other means or materials are described herein.

[0090]

[0090] When used herein, multiple elements, components, and / or materials may be presented in common lists for convenience. However, these lists should be interpreted as if each individual element of the list were individually identified as a distinct and unique element. Therefore, no individual component of such a list should be interpreted as being substantially equivalent to any other component of the same list based solely on its presentation in a common group, unless otherwise stated in the contrary. [Examples]

[0091]

[0091] Features of the present disclosure can be further illustrated by the following embodiments, but these embodiments are given for illustrative purposes only and should be understood not to limit the scope of the present invention unless otherwise stated. ●Example 1 - General Animal Feeding Behavior Modeling Protocol

[0092]

[0092] The following embodiments correlated load sensor data with specific feeding behaviors by collecting training data based on multiple studies conducted primarily using dogs, and then using that training data in additional studies to determine an acceptable model based on the training data. In these studies, the training and test data were correlated for accuracy using video recordings as ground truth data for comparison purposes. Several additional studies using cats were also conducted to verify that canine feeding behavior modeling could be performed in a similar manner to that for cats. Data processing in some embodiments included identifying feeding sessions (one or more) that included feeding segments in which the pet was interacting with a bowl (removing human interaction), dividing the feeding session or feeding segment into time increments, building features on top of the time increments, and using the features as input to a machine learning model that had ground truth video data for those time increments as output. For classification models (class-level outputs), the best model was selected typically using recall, precision, and / or f1 score. For continuous result models, MAE, MAPE, RMSE, R 2 The best model is determined using this method. ●Example 2 - Identification of human interaction

[0093]

[0093] Figure 12 illustrates how such data can be identified and trimmed from animal feeding data collected from a smart pet bowl equipped with one or more load sensors according to the present disclosure. Possible pre- and post-feeding human interactions can be initially established based on the proportion of the total duration of the collected dataset. For example, in this embodiment, 40% of the duration is used in the initial stage of the feeding session and 20% of the duration is used in the final stage of the feeding session. To more carefully identify the start and end of an actual feeding session and to remove human interactions, moving variances of load values ​​collected at fixed time values, e.g., 1 second, 2 seconds, 4 seconds, etc., can be used. In other words, possible pre-feeding segments and possible post-feeding segments can be checked incrementally within a fixed time frame, and peaks that are several times larger than the average peak occurring during feeding segments that are not identified as pre- or post-feeding segments can be identified. If a peak in one or more pre- or post-meal segments has an amplitude at least three times (or at least five times, for example) greater than the standard peak occurring between the estimated meal segment portions (or a standard peak of similar size found in possible pre- or post-meal segments), the signal may indicate human interaction. The human interaction closest to the estimated meal segment portion can be used as a point to trim from the available dataset, leaving identified meal segments without human interaction. This same or similar process can be used to remove other types of data that are not useful for collecting feeding behavior data. ●Example 3 - Data preprocessing and cleanup

[0094]

[0094] In addition to removing human interactions that can be identified and removed from the dataset, load sensor data can be preprocessed and cleaned up using modeling logic. For example, ground truth video data was compared with training data and also with test data from a dog population to identify video data that was inconsistent with the load sensor data and should be cleaned up from the dataset. More specifically, the collected load sensor data was cleaned to further clarify the relevant load sensor signals associated with various feeding behaviors, e.g., eating, dropping food, licking, and touching food with the nose. Using these feeding behaviors, load sensor microevents during the behavior (or individual loads applied during a microevent or duration) were collected based on threshold loads associated with the animal population being evaluated. For dogs in this study, the threshold was approximately 1 gram or greater than 1 gram. In this scenario, the logic was programmed to ignore any action if the microevent was less than 1 gram and other criteria were met indicating a low correlation between the load sensor data and the ground truth data collected by video recording.

[0095]

[0095] In addition, preprocessing of load sensor data may include identifying the start and end points of a full meal session (or the start and end points of smaller meal segments within a full meal session). Identifying the start and / or end points of meal segments may involve predicting feeding behavior (or actions) with the human interactions identified from the session, and mapping animal interactions at the timestamp level such that the first and / or last timestamps of feeding (or drinking) are considered the start and / or end points of the meal, respectively. In particular, in these embodiments, since meals are typically longer in duration, only small snapshots of the full meal are shown for illustrative purposes.

[0096]

[0096] In this embodiment, to determine the starting point, food was provided in a smart pet bowl, and the average baseline load value of the filled bowl was measured to determine the weight or mass of food provided before feeding began. Once the animal began eating, a variable load signal was collected, and the starting point was determined by the load sensor data. In some cases, the animal being evaluated began eating immediately after the pet food was provided, which did not give enough time to determine the average baseline load value. In these cases, the weight or mass of the food provided was calculated using other data collected throughout the session, which included collecting data during feeding interruptions, intake rates, time increments between shorter meal segments, etc.

[0097]

[0097] In determining the stopping point in this embodiment, one of two scenarios occurred, which included either the food being completely consumed or the animal stopping eating before the food was completely consumed. The food being completely consumed was determined when the load value of the bowl matched the load value known for the bowl before it was filled with food, e.g., the value when 0 grams of food remained. On the other hand, if there was food remaining in the smart pet bowl, a calculation was used based on the minimum possible value at the end of a fixed period of the session, e.g., the last 1 second, the last 3 seconds, the last 5 seconds, etc. ●Example 4 - Meal Session Interaction and Time Increment

[0098]

[0098] Figures 13 to 15 each show a portion of a single meal session. Figure 13 shows various discrete groupings of interactions (each grouping includes multiple micro-events, e.g., individual wrapping, individual licking, etc.) using boxes to indicate each interaction the animal has with the smart pet bowl or the pet food contained therein, based on weighted data. Figure 14 shows dotted lines indicating time increments that can be used to bundle the data. In this embodiment, time increments were bundled in increments of 0.075 seconds. Figure 15 combines discrete groupings of interactions and splits them to include the relevant time increments that occur between discrete interactions. Model performance is improved by splitting the data based on the above embodiment and finding features on the resulting time series. Different session / interaction / time window combinations were selected for different models to optimize model performance. ●Example 5 - Repetitive Feeding Modeling

[0099]

[0099] A study of dogs was conducted, which included evaluation of multiple features, 333 training sessions, and 67 animal evaluation sessions. Some of the features in this study included, to name a few, content activity duration removal, number of peaks in normalized loads, interaction duration, number of peaks, number of troughs per timestamp in normalized loads, number of peaks in gradients, autocorrelation of lag 2 to the RMS of gradients, intermediate gradient values, number of peaks in gradients, and autocorrelation of lag 3 to gradients.

[0100]

[0100] When processing data from the load sensor of the smart pet bowl of this disclosure, various machine learning models for continuous variables can be used. Mean absolute error (MAE), mean absolute percent error (MAPE), root mean square error (RMSE), coefficient of determination (R 2We identified the best model using continuous variables based on various machine learning logic systems employing ). The data is shown in Table 2 below. [Table 2]

[0101]

[0101] As can be seen in Table 2, R exceeds 90% 2 As indicated by the values, there was a strong linear relationship between the predicted number of repetitions and the actual number of repetitions for both the training and evaluation sessions. The mean absolute error in predicting the number of feeding repetitions was less than + / -2 repetitions for both the training and evaluation sets. ●Example 6 - Load signature for wrapping and licking

[0102]

[0102] Referring here to Figures 16 and 17, two sample datasets are shown containing load signatures that occurred while dogs were licking water from a smart pet bowl (lapping) (Figure 16) and while dogs were licking water from a smart pet bowl (licking) (Figure 17). This data represented short periods, i.e., a few seconds, which could vary from animal to animal. However, these load signatures, which show the peaks and troughs of the load, are distinguishable from one another, for example, by the more rhythmic pattern of lapping compared to the more disordered pattern of licking. These visual indicators helped in identifying appropriate features to distinguish between lapping and licking. For example, some of the identified features included peak height, distance between peaks, standard deviation of the distance between peaks, number of peaks, and peak and trough features. To summarize the examples of these two load signatures, wrapping appeared more rhythmic, in which the animal placed its tongue in the bowl, created a column of water, bit the column, and then the water fell back into the bowl. Licking, on the other hand, appeared more disorderly, in which the animal used its tongue to move it around the bottom of the bowl, collecting the water remaining on the surface. The two load signatures shown in this example are visually consistent with what is understood to be wrapping and licking, and as can be seen, they are shown to be distinct enough to be distinguished by their load signatures without the need for any other input. ●Example 7 - Processing of meal segment data

[0103]

[0103] In this evaluation, meal segment identification during a single session was assessed for meal duration (seconds) and meal intake (grams). As shown in Table 3A, values ​​were determined for the provided pet food, remaining pet food (after the meal), meal start time, and meal end time, and then, as shown in Table 3B, values ​​were determined using various regression error metrics. A machine learning model was constructed for meal segment identification, including i) continuously variable duration and ii) continuously variable intake. Mean absolute error (MAE), mean absolute percent error (MAPE), root mean square error (RMSE), and / or coefficient of determination. ( R 2 We studied ) and identified the best model. For example, MAE was found to be useful for understanding the mean deviation of predictions from actual values, and R 2 This was found to be useful for evaluating the linear relationship between actual values ​​and predicted values. [Table 3]

[0104]

[0104] As can be seen from the collected data, the results show that the estimated duration was approximately + / - 9 seconds on average of the true duration, with the true duration being approximately 222 seconds on average (+ / - 4%), and the estimated dietary intake was approximately + / - 6 grams on average of the true intake, with the true intake being approximately 193 grams on average (+ / - 3%). The relationship between predicted and observed values ​​for duration and intake, with an R^2 value of over 98%, indicates that the duration and intake models are very accurate. ●Example 8 - Classification of eating activities within a meal session (eating and / or drinking)

[0105]

[0105] The training protocol can be used to improve predictive logic related to various feeding and / or drinking behaviors. In this study, six feeding behavior parameters were evaluated, including feeding, pausing during feeding, touching the bowl, licking, touching food with the nose, and dropping food, and these were compared to predicted values ​​for individual micro-events. As described in Example 1, ground truth data in the form of video recordings were used in two different studies, one of which relates to training the smart pet bowl, e.g., training data, and the other studies relate to testing the smart pet bowl, e.g., test data. Various machine learning classifiers were used, for example. The training data was used to determine feeding behaviors, and the test data was used to identify acceptable classification models, which in this case were considered the best classification models. The actual data were compared to the predicted data using a confusion matrix along with recall, precision, and F1 score.

[0106]

[0106] Table 4 below shows some exemplary results demonstrating the high predictive power of the model that can be performed for classes with high levels of support. The model performs well in predicting feeding, pausing, contact with the bowl, and licking. Similar performance was not obtained for touching food with the nose and dropping food using the parameters selected above for this study, but this may be improved, for example, with different load sensitivities and / or additional training data. [Table 4]

[0107]

[0107] As shown in Table 4, the model performed very well in identifying feeding and pausing with an F1 score of over 90%. The model also performed well in identifying licking and bowl contact with an F1 score of over 80%. The model did not perform well for the less-supported classes of nasal contact and food dropping, but it can continue to improve with additional support, as the results were found to improve with additional data. However, these behaviors are extremely rare in canine feeding sessions. ●Example 9 - Modeling performance during drinking session

[0108]

[0108] In more detail regarding characterizing wrapping compared to licking during drinking sessions, various modeling or logic can be used to collect and process data from load sensors. For example, the use of machine learning logic can be used to collect data in relation to various drinking behaviors, e.g., wrapping, licking, and others, such as pausing (evaluating four classifications or groupings of drinking behaviors). In this embodiment, data was collected based on 190 test sessions where the time increment (or time window) used to collect data was set to 0.4 seconds per window. The results of this study are shown in Table 5 below. [Table 5]

[0109]

[0109] As shown in Table 5, the model performed very well in that it could identify wrapping, licking, or other behaviors with an F1 score of over 90%. The model also performed well in identifying licking with an F1 score of over 80%. ●Example 10 - Combined Modeling Using Both Wrapping and Licking Data

[0110]

[0110] A composite model that takes into account both wrapping and licking repetitions is also useful for modeling pet feeding behavior for the purpose of monitoring and / or discovering health insights. Table 6 shows data collected from a composite repetition model for wrapping and licking in dogs, along with the actual repetition (γ) distribution, and Table 7 shows model performance data. [Table 6] [Table 7]

[0111]

[0111] As can be seen from Tables 6 and 7, the model performed well, exhibiting a strong linear relationship between predicted and observed iterations, as indicated by values ​​exceeding 90% for both the training and test sets. The mean absolute error for the test set was above or below + / - 26.1 for 5 iterations (+ / - 19%). For this licking and wrapping combination model, the features that provided the best performance were i) interaction duration, number of load peaks, number of peaks relative to interaction duration, total number of normalized load peaks, and number of negative peaks relative to interaction duration. ●Example 11 - Iterative Modeling Using Normalization Logic

[0112]

[0112] The use of iterative models with normalization logic provides a way to merge data, such as merging the outputs from the behavior classification model and the composite iterative model to determine the iterations of wrapping and licking. Figure 18 shows exemplary load data collected during a drinking session, including both wrapping and licking (as well as two instances of contact with the bowl). Figure 19 shows how normalization logic can be applied to the data collected as shown in Figure 18 or other similar embodiments. ●Example 12 - Performance of a single feeding behavior model

[0113]

[0113] To illustrate the model performance using a separate iterative model for a single feeding behavior in dogs (not a combination of two or more feeding behaviors), i.e., wrapping, data was collected and processed as shown in Tables 8 to 10. 190 test sessions were used with this data, along with a machine learning model set to default parameters. In this example, Table 8 shows the data results when using a separate wrapping model with an iterative (γ) distribution, Table 9 shows the results for the wrapping model performance, and Table 10 shows the results for normalized wrapping performance. [Table 8] [Table 9] [Table 10]

[0114]

[0114] As can be seen from this data, the normalized wrapping output model in Table 10 performed better than the wrapping model in Table 9. ●Example 13 - Pet Health Insights

[0115]

[0115] With regard to health insights, various health issues can be correlated with feeding behavior data collected based on animal interaction with the smart pet bowl of this disclosure. Health insights obtainable from the smart pet bowl and related systems and methods may include identification of the animal's overall health, such as dental health, diabetes, kidney health, and digestive conditions. For example, changes in feeding behavior, such as an increase or decrease in feeding or drinking frequency, duration of feeding or drinking, feeding or drinking speed, repetition or pause in chewing or licking, and amount ingested, can provide insights into the animal's health. ●Example 14 - Collection and processing of cat load data

[0116]

[0116] Much of the data shown and described in the embodiments described above relates to studies of dogs. However, the feeding behavior of cats can be similarly collected using one or more load sensors associated with the smart pet bowl according to the present disclosure. The load data can be collected and processed in a similar manner to that described above with respect to studies of dogs. However, cats tend to eat and drink faster than dogs, and in addition, cats tend to eat and drink more delicately on average than dogs. Therefore, for example, in a smart pet bowl set up for use by cats, it may be useful to have load sensors (one or more) with somewhat enhanced sensitivity compared to the load sensor sensitivity used for dogs. However, excessive sensitivity of the load sensors may not be a particular concern if the load sensors (one or more) are sensitive enough to capture the microbehavior of the subject to be captured using the load sensor data and respond quickly enough.

[0117]

[0117] As an example of data collected from a study of cats, a smart pet bowl was used with four individual load sensors equidistant from each other along the bottom surface of the bowl. In particular, fewer or more load sensors, for example 1 to 10, can be used. In this example, four separate loads from the four load sensors were collected and then added together to approximate the load that could be obtained by a smart pet bowl with a single load sensor. Recorded data for total count, recorded average load, standard deviation, recorded minimum load, percentage of counts below a specific load value, and maximum load were recorded and evaluated. In this study, the total cumulative load from the load sensors recorded less than approximately 75 grams for 99.5% of the counts. From this cat data, the use of load sensors (one or more) with appropriate load sensitivity was selected.

[0118]

[0118] Specific embodiments of the present invention are disclosed herein. Specific terms are used, but these are general and illustrative only and are not intended to limit the scope. The scope of the present invention is defined in the "Claims." In light of the above-described art, many modifications and variations are possible to the present invention. Therefore, it is understood that the present invention can be implemented within the scope of the appended claims, even if not specifically described.

Claims

1. A smart pet bowl for monitoring pet feeding behavior, i) a pet bowl for holding pet food or water, or ii) a pet bowl insert for holding pet food or water, A load sensor capable of sensing changes in the load of pet food or water contained in the pet bowl and associated with the pet bowl, having a sensitivity of + / - 50 grams or less, and the load data that can be collected from the load sensor having a sampling rate of 10 to 150 samples per second in continuous time increments of approximately 0.01 seconds to 5 seconds, A data communication device that communicates the load data via a computer network, Processor and A memory that stores instructions for communicating the load data via the computer network when executed by the processor, A smart pet bowl equipped with [features / equipment].

2. The smart pet bowl according to claim 1, comprising a bowl support for housing the load sensor and a bowl insert for housing the pet food or water, and configured to transmit a load from the bowl insert to the load sensor.

3. The smart pet bowl according to claim 1, wherein the smart pet bowl is a single, integrated pet bowl configured to directly contain pet food or water, and the single, integrated pet bowl also houses the load sensor at a position where the animal's interaction with the pet food generates load data.

4. The smart pet bowl according to claim 1, wherein the sample rate, the continuous time increment, or both are controlled onboard by the smart pet bowl.

5. The smart pet bowl according to claim 1, wherein the sample rate, the continuous time increment, or both are controlled by a client device via the computer network.

6. The smart pet bowl according to claim 1, wherein the smart pet bowl is capable of filtering out human load data, false triggers, or accidental interactions with the smart pet bowl or its contents.

7. The smart pet bowl according to claim 1, wherein the smart pet bowl can distinguish between multiple pets in a household with multiple pets.

8. The smart pet bowl according to claim 1, wherein the sensitivity, sample rate, and continuous time increment are established at a level sufficient to identify count-based feeding behavior selected from wrapping, licking, or biting.

9. The smart pet bowl according to claim 8, wherein the sensitivity, sample rate, and continuous time increments are established at a level sufficient to allow for the counting of individual microevents of the feeding behavior within a single time increment or over a continuous period spanning multiple time increments, wherein the individual microevents include individual wrapping, individual licking, or individual biting actions.

10. The smart pet bowl according to claim 1, wherein the sensitivity, sample rate, and continuous time increment are established to a level sufficient to identify duration-based feeding behaviors selected from touching the pet bowl, moving the pet bowl, touching the pet food with the nose, pausing, feeding, wrapping, licking, or a combination thereof.

11. The smart pet bowl according to claim 10, wherein the sensitivity, sample rate, and continuous time increments are established at a level sufficient to enable continuous mapping of the time increments during which the duration-based feeding behavior occurs or does not occur.

12. The smart pet bowl according to claim 1, further comprising a secondary sensor selected from a proximity sensor, a camera, a microphone, an accelerometer, a gyroscope, an inertial measurement unit sensor, a radar, or a combination thereof.

13. The smart pet bowl according to claim 1, wherein the smart pet bowl is a water bowl for dogs, the load sensor has a sensitivity of + / - 4 grams or less, the sample rate is 15 to 75 samples per second, and the time increment is at least about 0.4 seconds.

14. The smart pet bowl according to claim 1, wherein the smart pet bowl is a food bowl for dogs, the load sensor has a sensitivity of + / - 4 grams or less, the sample rate is 15 to 75 samples per second, and the time increment is at least about 0.3 seconds.

15. The smart pet bowl according to claim 1, wherein the smart pet bowl is a water bowl for cats, the load sensor has a sensitivity of + / - 2 grams or less, the sample rate is 15 to 75 samples per second, and the time increment is at least about 0.4 seconds.

16. The smart pet bowl according to claim 1, wherein the smart pet bowl is a food bowl for cats, the load sensor has a sensitivity of + / - 2 grams or less, the sample rate is 15 to 75 samples per second, and the time increment is at least about 0.3 seconds.

17. A system for monitoring the feeding behavior of pets, Smart Pet Bowl and Processor and Memory for storing instructions, Equipped with, The aforementioned smart pet bowl is i) a pet bowl for holding pet food or water, or ii) a pet bowl insert for holding pet food or water, A load sensor capable of sensing changes in the load of pet food or water contained in the pet bowl and associated with the pet bowl, having a sensitivity of + / - 50 grams or less, and the load data that can be collected from the load sensor having a sampling rate of 10 to 150 samples per second in continuous time increments of approximately 0.01 seconds to 5 seconds, A data communication device that communicates the load data via a computer network, Includes, When the instruction is executed by the processor, Receiving the load data from the data communication device, Identifying feeding behaviors that occur within one or more of the time increments based on the pet's interaction with the pet bowl or its contents, A system that includes this.

18. The system according to claim 17, wherein the processor and the memory are mounted and arranged in the smart pet bowl.

19. The system according to claim 17, wherein the processor and the memory are physically located remotely from the smart pet bowl and communicate with the data communication device via a network.

20. The system according to claim 17, further comprising: the memory storing instructions; when the instructions are executed by the processor, the system excluding load data of interactions with the smart pet bowl or its contents if it is determined to be a human interaction, an incorrect trigger, or an accidental interaction.

21. The system according to claim 17, wherein the feeding behavior is a count-based feeding behavior selected from wrapping, licking, or biting.

22. The system according to claim 17, wherein the feeding behavior is a duration-based feeding behavior selected from touching the pet bowl, moving the pet bowl, touching the pet food with the nose, pausing, eating, wrapping, licking, or a combination thereof.

23. The system according to claim 17, further comprising the memory storing instructions and notifying the pet's caretaker of the feeding behavior or a change in the pet's feeding behavior when the instructions are executed by the processor.

24. The system according to claim 23, wherein notifying the administrator includes warning the administrator that the pet may have a health problem.

25. The system according to claim 17, wherein the smart pet bowl further includes secondary sensors including a proximity sensor, a camera, a microphone, an accelerometer, a gyroscope, an inertial measurement unit sensor, or a combination thereof.

26. The system according to claim 17, wherein the smart pet bowl is a water bowl for dogs, the load sensor has a sensitivity of + / - 4 grams or less, the sample rate is 15 to 75 samples per second, and the time increment is at least about 0.4 seconds.

27. The system according to claim 17, wherein the smart pet bowl is a food bowl for dogs, the load sensor has a sensitivity of + / - 4 grams or less, the sample rate is 15 to 75 samples per second, and the time increment is at least about 0.3 seconds.

28. The system according to claim 17, wherein the smart pet bowl is a water bowl for cats, the load sensor has a sensitivity of + / - 2 grams or less, the sample rate is 15 to 75 samples per second, and the time increment is at least about 0.4 seconds.

29. The system according to claim 17, wherein the smart pet bowl is a cat food bowl, the load sensor has a sensitivity of + / - 2 grams or less, the sample rate is 15 to 75 samples per second, and the time increment is at least about 0.3 seconds.

30. The system according to claim 17, wherein the system can distinguish between multiple pets in a household with multiple pets.

Images