Method and apparatus for remote analysis of monitored target parameters

Abstract

This invention relates to a method and apparatus for analysing vital signs or other target parameters of an observed animal or human (e.g., an elderly person) or object from a distance (e.g., RF sensing). It is proposed to apply a decision making process based on decision criteria for when to send a notification (e.g., alert) based on observed short-term time-series of sensing data preceding a specific triggering event (e.g., loss of the vital signal) and / or a recent long-term trend in daily activities or events, indicative of early and late signs that typically precede a triggering event (e.g., a specific health condition such as death). The decision making process may be trained e.g. by artificial intelligence algorithm(s) to identify early signs and / or late signs prior to the specific health condition.

Description

[0001] 2024PF80346

[0002] 1

[0003] Method and apparatus for remote analysis of monitored target parameters

[0004] FIELD OF THE INVENTION

[0005] This invention relates to wireless sensing systems for use in various environments (such as health, logistics or automotive).

[0006] BACKGROUND OF THE INVENTION

[0007] Due to rapid advancements in communication technology, multiple equipment and devices are now capable of communicating with one another inside a network, referred to as the Internet of Things (loT). While radio frequency (RF) signals are transmitted, reflected, obstructed, and dispersed by things such as buildings, furnishings, automobiles, and living beings, it is possible to gather relevant data from received RF signals such as presence, position, moving directions, speed, and vital signals. RF sensing, as opposed to traditional hardware sensors, provides consumers with low-cost and unobtrusive services. Furthermore, because RF signals are broadcast, they may be utilized not just to monitor multiple individuals but also to record changes in the environment across a vast region. With the use of RF technologies, end-users do not need to carry equipment or disrupt their daily routine. RF signals can be used to monitor macro and micro activity and to track persons, animals or objects.

[0008] In the past years, both professional and home lighting have moved increasingly towards (wireless) connected systems. RF sensing enables people or object detection by analyzing RF disturbances caused by them.

[0009] Moreover, passive RF sensing (where no active user efforts, such as charging batteries or wearing devices are required) allows monitoring vital signs or other signalmodulating parameters from a distance without involving the monitored object, animal or person.

[0010] However, sensing signal readings can be temporarily lost and may therefore trigger wrong decisions about a triggering event of a monitored object, animal or person. There is hence a need for a sensing system (e.g., a low-cost and / or privacy-preserving and / or non-wearable-device-based sensing system) for notifying monitoring staff (e.g., nursing staff)2024PF80346

[0011] 2

[0012] of a possible triggering event (e.g., possible death or other health condition) of an animal or person (e.g., a nursing home resident prior to anyone arriving in the elderly's room) or object.

[0013] US2021 / 057101A1 relates to in-home monitoring and an early health crisis alarm system for elderly individuals and patients with chronic diseases.

[0014] US2020 / 297955A1 relates a system for managing a chronic disease of a user such as a chronic respiratory or cardiac disease. The system may include a physiological monitor adapted to be carried by the user and operative to sense a physiological parameter of the user. The system may include a management device operatively coupled with the physiological monitor to receive the sensed physiological parameter of the user.

[0015] US12161461B2 relates to a method for detecting an undesirable event or condition in a patient. The method may comprise receiving a first input signal from an UWB radar configured to monitor an environment occupied by the patient and including information representative of the patient's motion. Data derived from the first input signal is processed using a pattern recognition model to detect and classify patterns in the data derived from the first input signal as indicative or predictive of an undesirable event or condition involving the patient.

[0016] SUMMARY OF THE INVENTION

[0017] It is an object of the present invention to provide a sensing system for reliable remote analysis of monitored target parameters.

[0018] This object is achieved by a method as claimed in claim 1, by an apparatus as claimed in claim 10, and by a computer program product as claimed in claim 14.

[0019] According to a first aspect, a method for improving reliability of detecting a triggering event at a monitored human, object or animal is provided, which comprises:

[0020] remotely sensing (e.g., comprising RF sensing) at least one target parameter of the monitored human, object or animal;

[0021] storing sensing data derived from the sensed at least one target parameter; and applying at least one decision criterion to both a short-term time-series and a long-term trend of the stored sensing data to decide about the detection of the triggering event;

[0022] wherein the at least one decision criterion comprises a first decision criterion applied to the short-term time series of the stored sensing data and a second decision criterion applied to the long-term trend of the stored sensing data.2024PF80346

[0023] 3

[0024] According to a second aspect, an apparatus for improving reliability of detecting a triggering event at a monitored human, object or animal is provided, wherein the apparatus comprises:

[0025] a sensing unit for remotely sensing (e.g., comprising RF sensing) at least one target parameter of the monitored human, object or animal;

[0026] a database for storing sensing data derived from the sensed at least one target parameter; and

[0027] a decision making module for applying at least one decision criterion to a short-term time-series and a long-term trend of the stored sensing data to decide about the detection of the triggering event;

[0028] wherein the at least one decision criterion comprises a first decision criterion applied to the short-term time series of the stored sensing data and a second decision criterion applied to the long-term trend of the stored sensing data.

[0029] According to a third aspect, a computer program product is provided, which comprises code means for producing the steps of the method of the first aspect when run on a computer device.

[0030] Accordingly, a remote-sensing-based alerting system with improved reliability is proposed to detect a triggering event (e.g., a critical health condition, such as death, of a patient or other target conditions of a human, animal or object). While RF sensing or other passive or active remote sensing options are capable of monitoring target parameters (e.g., vital signs of an elderly person or patient or other signals indicating target parameters) from a distance, detection of the triggering event (e.g., loss of vital sign or appearance of a triggering signal level) may be either caused by absence / appearance of a monitored signal (level) or by a (low-cost) sensing system struggling to acquire clean enough sensing signals for a decision algorithm extracting the target parameter(s). This problem is addressed by applying one or more decision criteria for when to send an alert based on observed time-series of stored sensing data preceding a possible triggering event and long-term trend(s) of the stored sensing data, that typically precede a true triggering event.

[0031] For instance, RF sensing may monitor vibrations of a machine or movement patterns of mechanical parts within a machine which may be indicative of the “health” of the machine. In rotating equipment, vibration may arise due to the dynamic forces generated as these machines operate. These forces can cause components to move back and forth or oscillate around their equilibrium positions. The complexity of this motion can vary; it can be a simple, repetitive motion, or it can be more chaotic, involving multiple frequencies and2024PF80346

[0032] 4

[0033] directions. A first continuous vibration pattern in a specific rotating equipment may be an early sign of impending failure, a second vibration pattern may be a late sign of failure while an actual failure of the machine materializes by the machine' s arm no longer being able to perform its intended macro movement.

[0034] According to a first option that can be combined with any one of the first to third aspects, a sensing algorithm used for the remote sensing may be reconfigured based on the at least one long-term trend of the stored sensing data.

[0035] According to a second option that can be combined with the first option or any one of the first to third aspects, the at least one decision criterion may comprise a first decision criterion applied to a short-term time series of the stored sensing data and a second decision criterion applied to a long-term time series of the stored sensing data.

[0036] According to a third option that can be combined with the first or second option or any one of the first to third aspects, the triggering event may be a health condition, in particular death, of the human, wherein the at least one target parameter comprises a first target parameter related to vital signs and the at least one decision criterion is applied to an observed time-series and / or a long-term trend of sensed vital signs, in particular heartbeat and / or breathing signals, to decide about the detection of the health condition.

[0037] According to a fourth option that can be combined with the third option or any one of the first to third aspects, the at least one target parameter may comprise one or more other target parameters, in particular large and / or small movements and / or posture and / or activity patterns, in addition to the first target parameter, wherein one or more further decision criteria are applied to an observed time-series and / or a long-term trend of the stored sensing data related to the one or more other target parameters to decide about the detection of the health condition. According to a fifth option that can be combined with the fourth option or any one of the first to third aspects, one or more decision criteria applied to the observed time series of vital signs and / or the other target parameters may be used to determine late signs of sub-categories comprising at least one of reduced circulation, increasing worry and anxiety, stopped eating and drinking, loss of consciousness, and changed breathing pattern, and / or one or more decision criteria applied to the long-term trend of the vital signs may be used to determine early signs of sub-categories comprising at least one of lack of interest in the surrounding world, low mood, increased sleep, newly added confusion, reduced physical ability, and decreased appetite.

[0038] According to a sixth option that can be combined with any one of the first to fifth options or any one of the first to third aspects, the at least one decision criterion may be2024PF80346

[0039] 5

[0040] based on sub-category-based scores allocated to the observed time-series and / or the longterm trend of the stored sensing data.

[0041] According to a seventh option that can be combined with any one of the first to sixth options or any one of the first to third aspects, the detection of the triggering event may trigger at least one of an alert to a monitoring person or a change of a care program of the monitored human.

[0042] According to an eighth option that can be combined with any one of the first to seventh options or any one of the first to third aspects, the at least one decision ciriterion may be applied by a supervised artificial intelligence, Al, model.

[0043] According to a ninth option that may be combined with any one of the first to seventh options or any one of the first to third aspects,

[0044] According to a tenth option that can be combined with any one of the first to ninth options or any one of the first to fourth aspects, the observed time-series of the stored sensing data may relate to early signs obtained from months up to a year or more before a triggering event, and the long-term trend of the stored sensing data may relate to late signs obtained weeks or days or even shorter before the triggering event.

[0045] According to an eleventh option that can be combined with any one of the first to tenth options or any one of the first to fourth aspects, the decision making module may be configured to access the database and / or a trend analysis module to obtain the observed timeseries and / or the at least one long-term trend of the stored sensing data for use in deciding whether the triggering event was falsely detected or not.

[0046] According to a twelfth option that can be combined with the eleventh option or any one of the first to third aspects, the trend analysis module may be configured to analyse the sensing data stored in the database to obtain a plurality of short-term scores and / or a plurality of long-term scores for use by the decision making module.

[0047] According to a thirteenth option that can be combined with any one of the first to twelfth options or any one of the first to third aspects, the decision making module may be configured to trigger a notification and / or an alarm by an alert unit in response to a decision that the triggering event was correctly detected.

[0048] According to a fourteenth option that can be combined with the thirteenth option or any one of the first to third aspects, the alert unit may be configured to trigger a message at a remote device or to initiate an optical and / or acoustical alarm at a remote station or office of the monitoring staff.2024PF80346

[0049] 6

[0050] It shall be understood that the method of claim 1, the apparatus of claim 11, and the computer program product of claim 15 may have similar and / or identical embodiments, in particular, as defined in the dependent claims.

[0051] It shall be understood that a preferred embodiment of the invention can also be any combination of the dependent claims or above embodiments with the respective independent claim.

[0052] These and other aspects of the invention will be apparent from and elucidated with reference to the embodiments described hereinafter.

[0053] BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In the following drawings:

[0055] Fig. 1 schematically shows a block diagram of a sensing device according to a first embodiment;

[0056] Fig. 2 shows a flow diagram of a reliable sensing and notification procedure according to a second embodiment; and

[0057] Fig. 3 schematically shows a block diagram of an interactive score-based analysing and decision making system according to a third embodiment.

[0058] DETAILED DESCRIPTION OF EMBODIMENTS

[0059] Embodiments of the present invention are now described based on sensing systems using RF sensing for remote monitoring of a human, object or animal. However, the present invention may also be used in connection with other wireless technologies that allow wireless transmission of sensing signals (e.g., vital signals) to a remote analysis device or system.

[0060] Throughout the following disclosure, remote sensing shall be understood to include passive remote sensing via sensing signals transmitted by a monitoring sensing device without any sensors applied to the monitored human, animal or object, and active remote sensing that includes sensors applied to the monitored human, animal or object and configured to transmit sensing signals via respective transmission links to the monitoring sensing device.

[0061] Furthermore, “early signs” shall refer to signs obtained from months up to a year or more before a triggering event, while “late signs”shall refer to signs obtained weeks or days or even shorter before the triggering event.2024PF80346

[0062] 7

[0063] It is further noted that throughout the present disclosure only those blocks, components and / or devices that are relevant for the proposed data distribution function are shown in the accompanying drawings. Other blocks have been omitted for reasons of brevity. Furthermore, blocks designated by same reference numbers are intended to have the same or at least a similar function, so that their function is not described again later.

[0064] The following embodiments are directed to a specific exemplary use case where passive RF sensing is used for monitoring vital signs from a distance in order to learn about a specific health condition (e.g., death) of a fellow resident of a nursing home or the like. This learning can be used to notify nursing staff of a possible or approaching death of the nursing home resident prior to anyone arriving in the elderly's room. Thereby, as an example, negative experience of entering a room to find a deceased person can be prevented. This is not only true for co-residents but also for unsuspecting nursing staff entering the room.

[0065] Embodiments provide a low-cost (possibly privacy-preserving) non-wearable-device-based sensing system for providing a notification or alert about a triggering event such as a specific health or mental or other condition of a surveilled object or human or animal (e.g., to detect and notify when a person in an elderly care home has passed away).

[0066] While RF sensing is capable of monitoring vital signs from a distance, there is a risk that a sensing signal (e.g., a vital sign such as a breathing or heartbeat signal) may be false, distorted or lost (e.g., by absence of the vital sign (i.e., death) or by the sensing system struggling to acquire clean enough RF sensing signals for an underlying detection or decision function (which may involve an artificial intelligence (Al) algorithm) that is used to extract the sensing signal. The RF sensing system may for instance struggle with wireless interference in the unlicensed band or a suboptimal positioning of the elderly's chest with respect to luminaires (of a lighting system) configured to perform the RF sensing. Sensing signal readings can be temporarily lost due to many different factors ranging from environmental factors (wireless traffic), certain movements of the monitored human, object or animal in adjacent rooms, and the specific location of a location of a sensing target (e.g., an elderly' s chest) with respect to the RF sensing arrangement. E.g., if a person is sleeping, the chest can remain for a prolonged period in an orientation not conducive for collecting vital signs by RF sensing. Movements can be large movements (e.g., gait) or small movements (e.g., trembling of hand). However, as passing away of an elderly person or other triggering events (e.g., stroke, stumbling, faint, epileptic seizure, earthquake, tsunami, etc.) of a target human, object or animal may come unexpected or may take long time, an alarm2024PF80346

[0067] 8

[0068] cannot be triggered every time the RF sensing system loses the sensing signal or receives a wrong sensing signal.

[0069] Therefore, in embodiments, decision criteria are set up and used for evaluating stored sensing data as to when to send an alert or notification. The evaluation may be based on sensing data stored prior to a triggering event (e.g., loss or actuation) of one or more specific sensing signals. For example, as described in embodiments, the stored sensing data may include at least one of observed time-series of RF sensing data (of e.g. vital signs) preceding the triggering event (e.g. sensing signal loss or activation) and a recent long-term data trend (e.g., daily activities or behavior of the monitored human, object or animal) indicative of early and / or late signs that typically precede a triggering event (e.g., death among older persons).

[0070] In the specific exemplary use case of monitoring the health condition of elderly people, a so-called “ageing in place” ideology has been developed, whereby older persons should be able to live at home for as long as possible and only most frail older persons in our society shall live in nursing homes. A current report of a Center for Decease Control and Prevention (CDC) showed that as many as 22% of people who passed away in recent years did so in a nursing home and that another 2% died in a hospice facility. In addition, 30% of the people passing away died in as an in-patient in an hospital. Post-mortem care is of utmost importance for both the staff of Long Term Care (LTC) homes as well as the co-residents. However, currently, experiences of staff and co-residents with post-mortem care are far more often negative than they are positive. From a co-residence' s perspective, the prevailing strategy of dealing with death in nursing homes and assisted living centers is silence. To be able to maintain the silence surrounding death requires that the nursing staff rather than other elder co-inhabitants discover first that an elderly person has passed away. Learning about death of fellow resident by walking into a room to find a deceased person is regarded the most negative experience in an elderly care home. This is not only true for coresidents but also for unsuspecting nursing staff entering the room. Nearing death, the older person’s breathing will become slower, irregular, shallow and wheezy. Moreover, older persons can hold their breath for a long time, which makes it difficult for sensing systems because every breath was believed to be the last. This process of irregular breathing can carry on for a prolonged time until death occurred and reduces the accuracy of breathing detection algorithms (e.g. RF sensing algorithms). As it is difficult to identify when the final stage of life begins, nurses currently regard dying as a happening, not a process; even if the dying elderly exhibits clear signs, nurses find it difficult (also given that they are very busy and2024PF80346

[0071] 9

[0072] overloaded) to identify early signs. Consequently, the actual passing away of a resident in a nursing home is often missed. However, for the experiences of both the co-residents and the nursing home staff it is of utmost importance to proactively manage the immediate aftermath of a resident's death. There is hence a need for a low-cost, possibly privacy-preserving, nonwearable-device-based sensing system for reliably notifying nursing staff of the possible death of an elderly person prior to anyone arriving in the room without warning.

[0073] Fig. 1 schematically shows a block diagram of a sensing device 10 according to a first embodiment, in which the health condition of a monitored elderly person 20 is monitored.

[0074] The sensing device may be mounted at a ceiling of wall of a surveilled room, e.g., as a local part of RF sensing-based alerting system, and may be configured to detect when the person 20 in an elderly care home has passed away.

[0075] The sensing device 10 comprises a wireless transceiver 12 (TRX, e.g., an RF unit) capable of receiving and / or transmitting a wireless transmission signal (e.g., an RF sensing signal 3) via a wireless interface (e.g., an antenna), a signal extraction module 13 (SE) capable of analysing a received signal obtained from the transceiver 12 to determine whether a specific sensing signal (e.g., modified / modulated component of the received signal) of a sensed target parameter (e.g., vital sign of the person) is included in the received signal. This may be achieved by at least one of decoding, demodulating, filtering, signal forming (e.g., pulse shaping) and analog-to-digital converting based on the characteristics of the sensing signal.

[0076] Furthermore, the signal extraction module 13 is configured to issue a trigger signal to a decision making module (DM) 16, when it determines that the received signal does not include the specific sensing signal (signal loss).

[0077] The extracted sensing signal is then stored in a database (DB) 14, e.g., an addressable memory, as a series of sensing data.

[0078] Additionally, a trend analysis module (TA) 15 may be provided that is configured to continuously or intermittently (e.g., periodically) access the database 14 and to identify predetermined time-dependent trends (e.g., variations and / or behaviors) of the stored sensing data over time, e.g., based on a pattern analysis.

[0079] In response to the receipt of the trigger signal from the signal extraction module 12, the decision making module 16 accesses the database 14 and / or the trend analysis module 15 to obtain long-term data and / or short-term data for use in deciding whether the detected signal loss was detected accidentally (false alarm) or a true absence of the monitored2024PF80346

[0080] 10

[0081] target parameter. The decision making process may be based predetermined decision criteria and / or a learning process that may involve artificial intelligence (Al) models / engines, as described later in more detail.

[0082] If the result of the decision making process at the decision making module 16 is a true loss of the monitored target parameter, a notification and / or alarm by an alert unit (AL) 17 is triggered by the decision making module 17. The alert unit 17 may be configured to trigger a message at a remote device (e.g., mobile device such as a laptop, tablet, smartphone or pager or the like) via a software application (app) or to initiate an optical and / or acoustical alarm at a remote station or office of the monitoring staff, e.g., via the transceiver 12 or another wired or wireless communication link (e.g., via a private or public network).

[0083] At least one of the signal extraction module 13, trend analysis module 15, decision making module 16 and alert unit 17 may be implemented by one or more computing units (e.g., a software-controlled microprocessors) capable of performing respective ones of the above mentioned data processing actions.

[0084] The above described modules 13 and 15 to 17 are thus configured to improve reliability of target parameter detection for notifying a triggering event (e.g., death) by combining one or more decision criteria, which may be based on observed time-series of sensing data preceding the triggering event (e.g., loss of the sensing signal) and / or recent long-term trend(s) in daily activities of the elderly person 20, indicative of early and late signs that typically precede the triggering event, wherein the early and late signs are derived from RF sensing via the transceiver 12.

[0085] We therefore describe decision criteria for when to send an alert based on (A) our observed time-series vital sign RF sensing data preceding the loss of sensing signal as well as (B) the recent long-term trend in daily activities of the elderly indicative of early and late signs that typically precede dying among older persons. While it is difficult for the physicians to predict dying among the older persons who did not have cancer, studies have shown that nurses are able to identify both manifest and subtle signs of dying. We therefore train Al algorithms to identify both “early signs” from several months up to a year before dying, as well as “late signs” observed days or weeks before dying.

[0086] Fig. 2 shows a flow diagram of a reliable sensing and notification procedure according to a second embodiment, which can be implemented (e.g., based on a software routine) in the computing units of the above first embodiment of Fig. 1, e.g., to enhance reliability of notifying the triggering event.2024PF80346

[0087] 11

[0088] In step S201 (MVS), vital signs of a monitored person (e.g., the elderly person 20 of Fig. 1) are montitored (e.g., by RF sensing via the transceiver 12 of Fig. 12 or by other remote passive or active sensing options).

[0089] Then, in step S202, the vital signs are extracted (EVS) from the received RF sensing signal (e.g., the the signal extraction module 13 of Fig. 1) and stored (SVS) in a database (e.g., database 14 of Fig. 1).

[0090] In step S203, the procedure checks (e.g., at the signal extraction module 13 of Fig. 1) whether a loss (SL) of a vital signal (e.g., heartbeat or breathing signal) has been detected. If not, the procedure jumps back to step S201 and continuous monitoring and archiving vital signals. Otherwise, if a signal loss has been detected in step S203, the procedure branches to step S204 where a short-term (ST) and / or long-term (LT) trend analysis (TA) is applied (e.g., by the trand analysis module 15 of Fig. 1) to the stored extracted vital signs (sensing signal data) to allow evaluation of reason(s) for the detected signal loss.

[0091] In the next step S205, the procedure checks (e.g., by the decision making module 16 of Fig. 1) based on results of the trend analysis whether one or more criteria for creating an alert (AC) or other type of alam or notification are met by the trend analysis. If not, the procedure jumps back to step S201 and continuous montitoring, archiving and analysing vital signals. Otherwise, if at least one criterion for the alert is determined to be met in step S205, the procedure branches to step S206 where the alert is initiated (e.g., at the alert unit 17 of Fig. 1).

[0092] Regulations require that assisted living or convalescent care facilities must remove the body of a resident from their premises immediately. Hence nursing home staff cannot wait with the removal until night hours. After deaths, co-residents prefer that bodies of dead coresidents were removed “on the quiet” as to not to upset them.

[0093] Thus, when the sensing system detects the triggering event (e.g., the monitored elderly person 12 has passed away), the room doors may be closed (automatically), the monitoring staff (e.g., nurse) can be notified, and the body can be wheeled discretely out the back on a gurney.

[0094] In addition, embodiments may be configured to utilize real-time RF sensing occupancy heatmaps of the elderly care home to manage removal of the resident's body.

[0095] In embodiments related to the proposed RF sensing-based alerting system, decision criteria are applied (e.g., by the decision making module 16 of Fig. 1) for when to send an alert (and when not to alert) upon loss of the vital sign sensing signal(s) by the RF2024PF80346

[0096] 12

[0097] sensing system. A corresponding decision engine may be configured to analyze one or both of early signs and late signs of the archived sensing data, that are known to typically precede a triggering event (e.g., dying among older persons in nursing homes). These early signs and late signs may be derived with an Al algorithm (that may be trained via supervised learning) from the RF sensing data collected during the period preceding the triggering event (e.g., current loss of the vital sign signals).

[0098] For instance, the decision engine may be configured to check in the past RF sensing data whether there have been any “Going into a bubble” early signs, which are indicative of the elderly person wanting to withdraw from the outside world and not caring about things to the same extent as before. Such early signs may be multidimensional, e.g., physical, psychological, social and existential. While those early signs are small and subtle, studies have shown that a sharp eye of a nurse (or a person that knew the older person well) can notice them. Inspired by the inferences made by sharp nurses, the activities of daily living data acquired by our RF sensing system in the room of the elderly over several months can be used to deduce whether the elderly person has exhibited a “Going into a bubble” behavior change.

[0099] Additionally, RF sensing data that was recently collected may be checked for “The body begins to shut down” late signs, which are indicative of the body starting to prepare for death.

[0100] Another known late sign that often precedes dying by days or weeks is that the older person starts falling. Thus, before sending an alert about a loss of a vital sign, RF sensing data may be analyzed for near-falls or falls of the monitored person.

[0101] Another late sign is that the monitored person stops taking medicines.

[0102] The check for early signs that are typically preceding dying in older persons may be applied over a period of several months up to a year.

[0103] In embodiments described herein, first time-series data (late signs) may be used for a first kind of detection algorithm for real-time detection of a triggering event (e.g., heartbeat detection to check when the heartbeat stops) and second long-term data (early signs) may be used for a (different) second kind of detection algorithm (e.g., gait detection, as poor gait is a sign known as preceding death) for detecting long-term signs typically preceding the triggering event (e.g., death). The first and second algorithms may both be configured to use data of the sensing system (e.g., RF sensing system). They may be computed separately e.g. by an edge Al, which a deployment of Al applications in devices throughout the physical world. It’s called “edge Al” because the Al computation is done near2024PF80346

[0104] 13

[0105] the user at the edge of the network, close to where the data is located, rather than centrally in a cloud computing facility or private data center.

[0106] Thus, depending on collected long-term-trend data, the sensing algorithm for real-time detection of the triggering event may be reconfigured to selectively decide about which of the available sensing algorithms to activate at a given time for a given triggering event.

[0107] For instance, in case of death as the triggering event, for a first person having already shown early signs of death, a heartbeat detection (RF) sensing algorithm may be optimized to almost never lose the heartbeat of the first person and will thus more reliably detect a loss of hearbeat. This means that the detection process of the first algorithm can be accelerated if the heart of the elderly stops beating. While this algorithm is now really good in detecting heartbeats, the sensing system configered to run this first death detection algorithm my not be able to concurrently look e.g. at the first person’s gait at the same time. Thus, if early signs of death have been detected, there may no longer be a need to track these early signs of death with the sensing system. It may thus be reconfigured to only concentrate on detecting heart stopping with maximum reliability.

[0108] For a second person showing no signs of death in the long-term trend sensing data, the sensing system may remain being configured to continue monitoring the long-term health trend (e.g., gait) while using a less reliable algorithm for detecting that the heart has stopped (as the second person is much less likely to die at the present stage). Such a sensing configuration allows continued monitoring of the gait to gather data on early signs of death.

[0109] Fig. 3 schematically shows a block diagram of an interactive score-based analysing and decision making system according to a third embodiment.

[0110] A trend analysis module (TA) 15 (which may correspond to the respective module of Fig. 1) is configured to analyse the stored sensing data to obtain n short-term scores (e.g., numeric values) SClsT-SCnsr (late signs) and / or m long-term scores SCILT-SCmLT (early signs), which are supplied to a decision making module 16 (which may correspond to the respective module of Fig. 1). The number n of short-term scores may be different from the number m of long-term scores.

[0111] The first category of long-term scores SCILT-SCIHLT may reflect “Going into a bubble” scores (early signs) and may comprise scores SC1LT-SC6LT on the following six subcategories:

[0112] 1. Lack of interest in the surrounding world

[0113] 2. Low mood2024PF80346

[0114] 14

[0115] 3. Increased sleep

[0116] 4. Newly added confusion

[0117] 5. Reduced physical ability

[0118] 6. Decreased appetite.

[0119] For example, if activities of stored daily-living sensing data indicate that the older person who previously had been social no longer appreciates visits or prefers fewer visits (e.g. by grandchildren), the respective “Going into a bubble” score SCILT and / or SC2LT is increased.

[0120] Similarly, if the stored sensing data indicates that older persons starts to prefer having their food in their room, the “Going into a bubble” score SCILT and / or SC2LT is increased.

[0121] Similarly, if the activities-of-daily-living sensing data indicates that an elderly resident has become more silent (indicative of wanting to be left alone), his / her “Going into a bubble” score SCILT and / or SC2LT is increased.

[0122] Or if the elderly person has showing less interest in such activities as exercise, watching television or is following sports results less and listening less to the radio less than before, his / her “Going into a bubble” score SCILT and / or SC2LT is increased.

[0123] Increased sleep is another known early sign preceding death of elderlies. There could be an increased desire to lie down and rest, for example wanting to he on the bed again and rest after the morning routine. From the stored sensing data, it can be derived whether it has become difficult to wake the older person up, either in the morning or during the rest of the day. Furthermore, a sleep detection sensing algorithm can be applied on the sensing data to deduce whether the older person has fallen asleep more often during the day, while sitting, for instance, in the dining-room or in an armchair. Or, the trend analysis module 15 may determine that an older person who has never before taken a rest after dinner suddenly feels a need to do so. For all those early signs, his / her respective “Going into a bubble” score SC3LT is increased.

[0124] Additionally, the trend analysis module 15 may analyze the elderly's activities of daily living data for signs of newly added confusion, which is an early sign preceding dying. This may be achieved by evaluating whether older persons suddenly begin to behave and express themselves in a different way than before, as if something not being right. As a result, the “Going into a bubble” score SC4LT is increased.

[0125] Furthermore, the sensing data may be analysed for indications of reduced physical ability of the elderly. The fact that the older person becomes weaker and has a2024PF80346

[0126] 15

[0127] greater tendency to fall is regarded by nursing staff as an early sign preceding dying. As a result, the “Going into a bubble” score SC5LT is increased.

[0128] Similarly, a change in the ability to perform daily activities because of a general decline in his function is regarded by nurses to be an early sign. To monitor this general decline, the trend analysis module 15 may check for a small change in a person’s pattern of movement through the room. For instance, the monitored person may have lost the ability to stand up without the aid of a mobility device. As a result, the “Going into a bubble” score SC5LT is increased.

[0129] Similarly, for an elderly person that hasn’t fallen at all, the sensing data may indicate (near-)fall events or “doddery” that are out of the ordinary for this specific elderly. As a result, the “Going into a bubble” score SC5LT is increased.

[0130] Moreover, decreased appetite is another sign preceding death. Many older persons decide to reduce their food intake because they do not want to live any more. The trend analysis module 15 may thus refer to sensing data indicating food / drink-preparation activities or eating / drinking related activities such as hand movements associated with handling cutlery. When an older person refuses food that they previously enjoyed, this can be interpret as a resignation. As a result, the “Going into a bubble” score SC6LT may be increased.

[0131] The second category of (late signs) refers to signs in the end of life, e.g., days or weeks before dying and are represented by the short-term scores SClsT-SCmsr and may comprise scores SC1ST-SC5ST on the following five sub-categories related to the body beginning to shut down:

[0132] 1. Reduced circulation

[0133] 2. Increasing worry and anxiety

[0134] 3. Stopped eating and drinking

[0135] 4. Loss of consciousness

[0136] 5. Changed breathing pattern.

[0137] Anxiety often manifests as anger and frustration. Older persons are often less able to talk at this stage, they try to express themselves through body language, e.g. waving their arms or making a noise. Another late sign that can be sense is that an older person has a broader worry which manifests as restlessness, incoherent talk and hallucinations. The trend analysis module 15 may be configured to run specialized algorithms on the stored sensing data to derive inferences on these signs of anxiety.2024PF80346

[0138] 16

[0139] Similar to the above long-term scores, respective ones of the short-term scores SCI ST-SC5ST are increased based on the result of trend analysis for the above five subcategories.

[0140] Based on the scores received from the trend analysis module 15, the decision making module 16 decides about validity of the triggering event and issuance of an alert. This may be achieved by comparing an accumulated total score of one or more or all shortterm and / or long-term sub-categories with one or more respective thresholds or by using the long-term and / or short-term sub-category scores as input to a decision algorithm that reflects the impact of individual long-term and / or short-term sub-categories on the decision.

[0141] In embodiments, an Al-based decision engine may be implemented e.g. in the decision making module 16 and / or the trend analysis module 15 of Figs. 1 and 3 to provide a predictive Al-based system.

[0142] The stored sensing signal data may be input into an Al or Machine Learning (ML) model. This model may be trained (e.g., by supervised training with explicit feedback) with (statistically) prepared sensing data that reflects long-term and / or short-term trend related to a triggering event (e.g. death or other triggering event mentioned herein). By leveraging the capabilities of AI / ML, the sensing and alert system may dynamically adjust and refine its decision making criteria for improved reliability. The system may include a feedback mechanism to continuously update and improve the AI / ML model based on collected real-time data.

[0143] In various embodiments, the learning data for the Al may be derived from statistical and / or measured sensing data reflecting long-term and / or short-term sensing data of target parameters (e.g., vital signs, environmental parameters, etc.) related to the monitored condition prior to a triggering event (e.g., loss of a critical signal, occurrence of a critical signal) at a monitored human, object or animal.

[0144] In supervised learning, the Al model may be trained using a labeled dataset (e.g., time series of sensing data and / or sub-category scores), where each data point is tagged with the correct output (e.g., triggering event happened or not). The Al model learns to predict the output for new input data by analyzing the training data. The learning process may thus involve the steps of gathering time series of labeled training data), splitting the dataset into training, validation, and test sets, determining types of sensing signal data (e.g., sensing parameters such as specific vital signs or sub-categories) that will be used for training, and choosing a suitable algorithm for the Al model, such as support vector machines (SVM), decision trees, or neural networks. Training of the Al model may then be achieved by2024PF80346

[0145] 17

[0146] executing a selected algorithm on the training dataset. Then, the model's accuracy may be tested by using a test set of input data with known output.

[0147] In embodiments related to elderly care, the RF sensing system (e.g., the decision making module 16 and / or the alert unit 17) may generate a notification to recommend the care staff when to transition from routine care to end-of-life care.

[0148] To summarize, a method and apparatus for analysing vital signs or other target parameters of an observed animal or human (e.g., an elderly person) or object from a distance (e.g., RF sensing) have been described. It is proposed to apply a decision making process based on decision criteria for when to send a notification (e.g., alert) based on observed shortterm time-series of sensing data preceding a specific triggering event (e.g., loss of the vital signal) and / or a recent long-term trend in daily activities or events, indicative of early and late signs that typically precede a triggering event (e.g., a specific health condition such as death). The decision making process may be trained e.g. by artificial intelligence algorithm(s) to identify early signs and / or late signs prior to the specific health condition.

[0149] It is noted that the invention is not limited to RF-sensing-based montoring of vital of the above embodiments. It can be applied to various types of UEs or terminal devices, such as mobile phone, vital signs monitoring / telemetry devices, smartwatches, detectors, vehicles (for Vehicle-to-Vehicle (V2V) communication or more general Vehicle-to-everything (V2X) communication), V2X devices, Internet of Things (loT) hubs, loT devices, including low-power medical sensors for health monitoring, medical (emergency) diagnosis and treatment devices, for hospital use or first-responder use, Virtual Reality (VR) headsets, etc.

[0150] Moreover, the proposed long-term and / or short-term analysis of monitored condition(s) may comprise other triggering events (e.g., stroke, stumbling, faint, epileptic seizure, earthquake, tsunami, etc.) of a target human, object or animal and may be based on other suitable target parameters that can be sensed by a remote (RF) sensing system. The triggering event may as well be a sudden appearance of a specific (level of) sensing signal that requires triggering an alert. In the example of a monitored object, long-term and / or shortterm sensing data of physical parameters (movement, vibration, temperature, environment, weather etc.) may be used to evaluate an approaching earthquake or tsunami or the like.

[0151] Other variations to the disclosed embodiments can be understood and effected by those skilled in the art in practicing the claimed invention, from a study of the drawings, the disclosure and the appended claims. In the claims, the word “comprising” does not exclude other elements or steps, and the indefinite article “a” or “an” does not exclude a2024PF80346

[0152] 18

[0153] plurality. A single processor or other unit may fulfil the functions of several items recited in the claims. The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage. The foregoing description details certain embodiments of the invention. It will be appreciated, however, that no matter how detailed the foregoing appears in the text, the invention may be practiced in many ways, and is therefore not limited to the embodiments disclosed. It should be noted that the use of particular terminology when describing certain features or aspects of the invention should not be taken to imply that the terminology is being re-defined herein to be restricted to include any specific characteristics of the features or aspects of the invention with which that terminology is associated. Additionally, the expression “at least one of A, B, and C” is to be understood as disjunctive, i.e., as “A and / or B and / or C”

[0154] A single unit or device may fulfil the functions of several items recited in the claims. The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage.

[0155] The described operations like those indicated in the above embodiments (e.g., Fig. 2) may be implemented as program code means of a computer program and / or as dedicated hardware of the related network device or function, respectively. The computer program may be stored and / or distributed on a suitable medium, such as an optical storage medium or a solid-state medium, supplied together with or as part of other hardware, but may also be distributed in other forms, such as via the Internet or other wired or wireless telecommunication systems.

Claims

2024PF8034619CLAIMS1. A method for improving reliability of detecting a triggering event at a monitored human (20), object or animal, comprising:remotely sensing (S201), optionally comprising radio frequency, RF, sensing, at least one target parameter of the monitored human (20), object or animal;storing (S202) sensing data derived from the sensed at least one target parameter; andapplying (S204, S205) at least one decision criterion to both a short-term timeseries and a long-term trend of the stored sensing data to decide about the detection of the triggering event;wherein the at least one decision criterion comprises a first decision criterion applied to the short-term time series of the stored sensing data and a second decision criterion applied to the long-term trend of the stored sensing data.

2. The method of claim 1, further comprising reconfiguring a sensing algorithm used for the remote sensing based on the at least one long-term trend of the stored sensing data.

3. The method of any one of the preceding claims, wherein the triggering event is a health condition, in particular death, of the human (20), and wherein the at least one target parameter comprises a first target parameter related to vital signs and the at least one decision criterion is applied to an observed time-series and a long-term trend of sensed vital signs, in particular heartbeat and / or breathing signals, to decide about the detection of the health condition.

4. The method of claim 3, wherein the at least one target parameter comprises one or more other target parameters, in particular large and / or small movements and / or posture and / or activity patterns, in addition to the first target parameter, wherein one or more further decision criteria are applied to an observed time-series and the long-term trend of the2024PF8034620stored sensing data related to the one or more other target parameters to decide about the detection of the health condition.

5. The method of claim 4, wherein one or more decision criteria applied to the observed time series of vital signs and / or the other target parameters are used to determine late signs of sub-categories comprising at least one of reduced circulation, increasing worry and anxiety, stopped eating and drinking, loss of consciousness, and changed breathing pattern, and / or wherein one or more decision criteria applied to the long-term trend of the vital signs are used to determine early signs of sub-categories comprising at least one of lack of interest in the surrounding world, low mood, increased sleep, newly added confusion, reduced physical ability, and decreased appetite.

6. The method of claim 5, wherein the at least one decision criterion is based on sub-category -based scores allocated to the observed time-series and / or the long-term trend of the stored sensing data.

7. The method of any one of claims 3 to 6, wherein the detection of the triggering event triggers at least one of an alert to a monitoring person or a change of a care program of the monitored human (20).

8. The method of any one of the preceding claims, wherein the at least one decision ciriterion is applied by a supervised artificial intelligence, Al, model.

9. The method of any one of the preceding claims, wherein the observed timeseries of the stored sensing data relate to early signs obtained from months up to a year or more before a triggering event, and wherein the long-term trend of the stored sensing data relates to late signs obtained weeks or days or even shorter before the triggering event.

10. An apparatus for improving reliability of detecting a triggering event at a monitored human (20), object or animal, comprising:a sensing unit (12) for remotely sensing at least one target parameter of the monitored human (20), object or animal;a database (14) for storing sensing data derived from the sensed at least one target parameter; and2024PF8034621a decision making module (16) for applying at least one decision criterion on a short-term time-series and a long-term trend of the stored sensing data to decide about the detection of the triggering event;wherein the at least one decision criterion comprises a first decision criterion applied to the short-term time series of the stored sensing data and a second decision criterion applied to the long-term trend of the stored sensing data.

11. The apparatus of claim 10, wherein the decision making module (16) is configured to access the database (14) and / or a trend analysis module (15) to obtain the observed time-series and / or the at least one long-term trend of the stored sensing data for use in deciding whether the triggering event was falsely detected or not.

12. The apparatus of claim 11, wherein the trend analysis module (15) is configured to analyse the sensing data stored in the database (14) to obtain a plurality of short-term scores and / or a plurality of long-term scores for use by the decision making module (16).

13. The apparatus of any one of claims 10 to 12, wherein the decision making module (16) is configured to trigger a notification and / or an alarm by an alert unit (17) in response to a decision that the triggering event was correctly detected.

14. A computer program product comprising code means for producing the steps of any one of claims 1 to 9 when run on a computer device.

Images