GB2602668A

GB2602668A - Monitoring system

Info

Publication number: GB2602668A
Application number: GB2100367.8A
Authority: GB
Inventors: Moorhead Paul
Original assignee: Kraydel Ltd
Current assignee: Kraydel Ltd
Priority date: 2021-01-12
Filing date: 2021-01-12
Publication date: 2022-07-13
Also published as: GB202100367D0

Abstract

An arrangement of monitoring an environment (14, Fig. 2) involves using sensors (12A-H, Fig. 2) to monitor the environment, including a person in the environment, and detecting signs of life from the sensor data. Depending on the detection of signs of life, the value of an intervention parameter is adjusted. The intervention parameter value may decay over time towards the threshold TV1 or TV2, and be counter-adjusted in response to detection of signs that the person in the environment is well. Should the value reach a threshold TV1 or TV2, interventional action is taken to contact or otherwise check on the person in the environment. Interventional action may be in the form of at least one alert. The alert may involve sending an electronic message, a visual/audio signal and/or initiating a telephone or video call. The purpose of the invention is to provide a monitoring system capable of monitoring human behaviour and well-being in a similar manner and at a similar speed to that which a dedicated human carer could achieve.

Description

Monitoring System

Field of the Invention

The present invention relates to monitoring systems. The invention relates particularly to systems for 5 monitoring human behaviour and/or well-being in domestic environments.

Background to the Invention

Care for elderly, sick and vulnerable people is expensive to provide both for public or private care providers, and few people will be able to afford dedicated personal care to ensure that assistance is 10 provided immediately a problem arises.

Advances in low-cost computing and sensor technology, along with developments in machine-learning making it possible to monitor the home environment of vulnerable people at low cost.

It would be desirable to provide a monitoring system capable of monitoring human behaviour and well-being in a similar manner and at a similar speed to that which a dedicated human carer could achieve.

Summary of the Invention

From a flrst aspect, the invention provides a method of monitoring an environment using at least one sensor, the method comprising: monitoring said environment using said at least one sensor and generating corresponding sensor data; analysing said sensor data to detect at least one sign that is indicative of life in said environment; providing at least one intervention parameter for a person in said environment; taking at least one interventional action depending on a value of said at least one intervention parameter; incrementally adjusting the value of said at least one intervention parameter towards at least one threshold value that causes said at least one interventional action to be taken; and counter-adjusting the value of said intervention parameter to defer said at least one interventional action in response to detection of at least one relevant sign.

Typically. said monitoring involves monitoring at least one parameter of said environment using said at least one sensor and generating corresponding sensor data.

Preferably, the method includes comparing the value of said intervention parameter to at least one action threshold value, and taking said at least one action depending on the comparison.

Preferably, said counter-adjusting involves counter-adjusting the value of said intervention parameter away from said at least one threshold value or resetting the intervention parameter value to a limit value.

In preferred embodiments. taking at least one action involves generating at least one alert.

Generating said at least one alert may involves sending at least one electronic message and/or rendering at least one visual and/or audio signal and/or initiating a telephone or video call.

In preferred embodiments. said at least one relevant sign comprises at least one sign that is indicative of normal behaviour in said environment and/or of a normal condition of a human in said environment.

Preferably, said counter-adjusting involves adjusting the value of said intervention parameter in response to detection of each relevant sign, preferably as each relevant sign is detected.

Said counter-adjusting may involve adjusting the value of said intervention parameter by an amount or in a manner that is determined by a type of said at least one relevant sign.

Said counter-adjusting may involves adjusting the value of said intervention parameter by an amount 20 or in a manner that is determined by a confidence level with which said at least one relevant sign is detected.

Said counter-adjusting may involve adjusting the value of said intervention parameter by an amount or in a manner that is determined by a confidence level with which said at least one relevant sign is 25 deemed to indicate the person's well-being.

Optionally, said intervention parameter has a limit value. said intervention parameter value preferably being initialised to said limit value.

Typically: said adjusting involves adjusting the value of said intervention parameter at an adjustment rate. Said adjustment rate is optionally calculated based on a desired time taken for said counter-adjusting to adjust said intervention parameter to reach said at least one threshold value. Alternatively, said adjustment rate may be fixed.

Typically. said adjusting is performed between successive instances of said counter-adjusting.

Typically, said adjusting is stopped when said at least one relevant sign is detected.

In preferred embodiments: the method includes adjusting the value of said intervention parameter 40 value to expedite taking said at least one interventional action in response to detecting at least one abnormal sign that is indicative of abnormal behaviour and/or an abnormal human condition, and/or in response to detecting an absence of said at least one relevant sign, wherein adjusting the value of said intervention parameter value to expedite taking said at least one interventional action may involve adjusting the value of the intervention parameter value towards said at least one threshold value, or setting the value of the intervention parameter to said at least one threshold value.

Optionally, said adjusting involves the value of said intervention parameter in response to detection of each abnormal sign, preferably as each abnormal sign is detected.

Optionally, said adjusting involves adjusting the value of said intervention parameter by an amount or 10 in a manner that is determined by a type of said at least one abnormal sign.

Optionally, said adjusting involves adjusting the value of said intervention parameter by an amount or in a manner that is determined by a confidence level with which said at least one abnormal sign is detected.

Optionally, said counter-adjusting involves adjusting the value of said intervention parameter by an amount or in a manner that is determined by a confidence level with which said at least one abnormal sign is deemed to indicate the person's well-being.

In typical embodiments, said at least one sensor comprises one or more instances of any one or more of the following sensors: audio sensor; electro-optical sensor; image sensor; motion sensor; heart rate sensor; respiration sensor; ambient temperature sensor; pressure sensor and/or switch sensor.

Typically. said at least one parameter of the environment comprises any one or more of the following parameters: sounds; speech; light; movement; heart rate; respiration rate; temperature; use of equipment or objects.

Typically, said at least one sign may include any one or more of: detection of the person's respiration or heart rate (optionally including a determination that the respiration rate or heart rate is normal); movement in the environment; use of an appliance; preparation of food or drink; using the bathroom; turning a light on or off; non-verbal human utterances; opening or closing a door or window; talking, or making or receiving a phone call.

Optionally, said at least one intervention parameter is implemented by at least one timer.

Said analysing may involve comparing said sensor data with reference data representing detectable signs.

Said analysing may involve classifying said sensor data to detect said at least one sign.

Methods embodying the invention may be computer-implemented and/or implemented in hardware as applicable.

From a second aspect, the invention provides a monitoring system for monitoring an environment. 5 the system comprising; at least one sensor for monitoring said environment and generating corresponding sensor data; analysing means for analysing said sensor data to detect at least one sign that is indicative of life in said environment; means for providing at least one intervention parameter for a person in said environment; means for taking at least one interventional action depending on a value of said at least one intervention parameter; means for incrementally adjusting the value of said at least one intervention parameter towards at least one threshold value that causes said at least one interventional action to be taken; 15 and means for counter-adjusting the value of said intervention parameter to defer said at least one interventional action in response to detection of at least one relevant sign.

Systems of the invention may comprise any suitable means for performing any aspect of the method 20 described herein. Such means may include computer software and/or hardware as applicable.

Embodiments of the invention may be computer-implemented, or electronically implemented, or implemented using a combination of computing and electronic device(s) as is convenient.

The word "sign" as used herein is intended to embrace any activity, event, action, signal or condition relating to the monitored environment (including one or more person in the monitored environment as applicable) that is detectable using the provided sensor(s). Some signs may be directly detectable by one or more sensor. Other signs may be indirectly detectable, or may be inferred, from data captured by one or more of the sensors. Some signs are an indication that the person in the environment is well, while other signs may be an indication that the person is not well.

In typical embodiments, the system maintains at least one intervention parameter and determines whether not to take any interventional action depending on the value of the, or each, intervention parameter. Advantageously, in response to detecting a sign that the person is well, one or more intervention parameter value is adjusted or maintained to defer interventional action. Preferably, in response to detecting a sign that the person is not well, one or more intervention parameter value is adjusted to expedite interventional action. Typically, the or each interventional parameter value is incrementally adjusted over time so that, over time and in the absence of one or more signs that the person is well being detected, it reaches at least one threshold value in response to which interventional action is taken.

The intervention parameter may be a well-being parameter in that its value is an indication of a level of confidence that the person is well. As such its value may be maintained, increased or reset to an initial value (or other nominal value) in response to detecting a sign that the person is well. The parameter value may be decreased over time or in response to detecting a sign that the person is not well.

The intervention parameter may be a risk parameter in that its value is in indication of a level of risk that intervention is required. As such its value may be maintained, reduced or reset to an initial value in response to detecting a sign that the person is well. The parameter value may be increased over 10 time or in response to detecting a sign that the person is not well.

The intervention parameter may be implemented by a timer. The timer value may be an indication of a level of confidence that the person is well. As such the timer may be a countdown timer having a value that decreases over time. The counter value may be maintained, increased or reset in response to detecting a sign that the person is well, and may be decreased in response to detecting a sign that the person is not well. Alternatively, the timer value may be an indication of a level of risk that intervention is required. As such the timer may be a count-up timer having a value that increases over time. The counter value may be maintained, decreased or reset in response to detecting a sign that the person is well, and may be increased in response to detecting a sign that the person is not well.

In preferred embodiments, sensors, preferably including one or more sensors (e.g. optical sensor(s) and/or audio sensor(s)) that correspond to human capabilities such as hearing and sight, are used to detect and/or monitor one or more sign (typically an activity, event, action or condition, e.g. a physiological condition) relating to the environment and/or a person in the environment in order to determine behaviour(s), changes in behaviour(s), and/or the absence of behaviour(s) of the person, which in turn are possible indicators of the need for an intervention.

For example, detectable signs that are indicative of human condition or behaviour in an environment 30 (e.g. a domestic environment such as a home) include (but are not limited to): 1) A person's breathing or heartbeat 2) Movement around the home 3) Use of the TV set or other appliance 4) Preparing food or a drink 5) Turning a tap on or off 6) Using the toilet 7) Turning a light on or off 8) Non-verbal human utterances, e.g. coughing sniffing, sneezing, clearing a throat, laughter, crying, yells 9) Opening or closing a door or window 10) Talking, or making or receiving a phone call Such events or actions, and other signs relating to the environment or the person in the environment.

may be regarded as signs of life or well-being. Advantageously, sensor data corresponding to detected signs is used to update a intervention parameter for the person. The value of the intervention parameter may be used to determine whether or not action needs to be taken in relation to the person.

Some detectable actions, events or other signs may be regarded as proof, or close to proof, of the person's well-being being "good", e.g. detection of a healthy heart and/or healthy respiration rate.

Other activities, events or signs are weaker signs of good well-being, and/or may not be possible to detect with 100% accuracy. For example, detection of a cough might be confused by the sound of a nearby dog barking. It is preferred therefore to assign at least one confidence level, e.g. a probability value, to each detected sign for indicating the confidence with which the activity has been correctly identified and/or the confidence with which the detected sign is considered to be an indication of good well-being. Advantageously, the, or each, confidence level may be used to determine how the intervention parameter is updated.

The absence of sensor data or detected signs over time results in a reduction the confidence level that the person is well. Advantageously, the value of the intervention parameter is adjusted to reflect the reduction in confidence, in particular the actual reduction in confidence that the person remains safe/well given the duration of the period without data indicating that a reassuring sign has been detected. When the confidence level drops below a relevant threshold, an alert may be generated. In response, a human carer may visit, or otherwise contact, the person in order to verify that there is no problem or to provide any necessary assistance. Alternatively, an automated system may prompt the person to confirm that they are "well" before an alert is generated, and if they do so confirm, that can set the intervention parameter back to a value that indicates good well-being.

By way of example, in the context of a care home or other residence with multiple occupants in separate units, it might require many carer staff to visit the residents on a regular basis to ensure their well-being, and during the hours of darkness this could be very disruptive to the residents, particularly those with dementia. Advantageously, however, a intervention parameter value can be maintained for each resident based on sensor data associated with each resident. Any given resident may warrant a physical visit only if the confidence parameter value reaches a relevant threshold level. Thus fewer human resources may be required to monitor the residence.

Preferred embodiments of the invention are configured to use at least one sensor to monitor an environment (preferably including a person located in the environment, especially a person who lives alone) in order to detect signs, or signals, from the environment (including from the person, as applicable) from which the system can make inferences about the person's well-being, including for example the person's physical health, current activities and/or state of mind. The detectable sign may be or relate to biological or physiological signs such as heart-rate or respiration of the person.

The sign may be or relate to an activity or event e.g. moving, turning on the TV or operating other appliance(s), using a telephone: flushing a toilet: boiling a kettle, opening a door and so on. The sign may be a sound (or a smell) or something more abstract, e.g. a motion of air suggesting that a window has been opened, or a sudden change in light level suggesting that a light has been turned on/off or curtains opened or closed. The detectable signs may be regarded as "signs-of-life" or "signs-of-well-being".

Data from one or more sensor may be combined to make the inferences, for example a doorbell detected during the daytime, but no subsequent motion or door opening, suggests that the person is not responsive. Some signs are detectable by any combination of one or more conventional sensors (e.g. any one or more of motion sensors, light sensors, audio sensors, optical detectors, image sensors, pressure sensors, switch sensors and/or physiological sensors), although embodiments of the invention may be used with any sensor, e.g. sensors for detecting brain activity gaze and/or expression.

The signs and the inferences from them allow the preferred system to assign a "confidence level" value to a belief that the person is safe and well. For example, inferring that the person made a cup of tea and sat down to watch TV allows the system to set the confidence level to a high value (perhaps 100%, or 10 out of 10, or any other scale), whereas detecting, say, a cough or sneeze with, say, 80% accuracy, may cause the system to set the confidence level to 80% (assuming that the current confidence level is lower than that, since it is preferred not to reduce the confidence level from a higher value to a lower value just because a sneeze or the like is detected).

As time passes, in the absence of relevant signs being detected and inferences being made from them, confidence that the person is well may reduce, and so the preferred system models this as a "decay" or "decrease" in the confidence level with the passing of time. For example the confidence level may be decreased linearly at a fixed rate or a variable rate, e.g. the rate of decrease in confidence may increase as time passes without detection of a relevant sign. The decay profile may be arbitrary and may be adjusted to achieve a desired timing for any intervention(s) or other triggered action(s).

Without detection of a relevant sign, at some point, the confidence level reaches a threshold value at which the system may be configured to perform one or more actions, for example any one or more of: 1. Initiate a prompt within the environment (e.g. using TV or speaker or other device) or otherwise send a message to the person to ask the person if they are ok. If they answer "yes" by voice, or pressing a button or some other action, then the system may reset the confidence level to 100%. If they answer "no", the confidence level may be set to zero (or other lower value) immediately, and/or a carer may be alerted.

2. Alert a carer.

3. Initiate a video or phone call between the person and a remote carer.

More than one threshold may be supported, for example: a warning threshold at which a carer is prompted to check-in with the person within a given time period; and/or an alarm threshold at which a carer is instructed to check-in with the person immediately. The check-in might be in person, or by 5 telephone or video call, or by remote viewing of a video feed of the environment.

Some signs are negative, e.g. the aforementioned doorbell ring which is not followed quickly by motion, or telephone ringing without subsequent motion or conversation, may be considered to be negative signs with respect to well-being. When negative signs are detected, the system may decrease the confidence level; e.g. to a pre-defined lower value, or by decrementing it by a fixed amount.

Other approaches to "decaying confidence in well-being" may be adopted; for example including: * Inverting the measuring scale to indicate risk rather than confidence in well-being. For example 100% confidence in the person being safe and well may correspond to "zero risk" and so a risk level may start at zero. Over time; without detected signs; the risk level increases, and when it reaches a threshold, the system may take action in the same manner as described above. Detection of signs may serve to reset the risk to zero, or set it to a lower value, or potentially to higher values (for negative signs).

* An "intervention timer". For example, the system may be configured to allow no more than a specified period of time X (e.g. 120 minutes) to pass without reassuring signs before action is taken. A timer may starts from zero (or other base value). When a reassuring sign is detected, the timer may be reset (e.g. If the sign indicates 100% assurance of well being) or adjusted by a given amount (e.g. decremented by Y (say 30 minutes)) if the sign provides only partial reassurance.

All three approaches (confidence, risk, timer) can be tuned by (a) adjusting the rate at which the value increases/decreases and (b) the threshold levels. This allows the system to be configured to correspond to the relative frailty or level of concern about the person, and/or the level of monitoring 30 service being provided.

Further advantageous aspects of the invention will be apparent to those ordinarily skilled in the art upon review of the following specific embodiment and with reference to the accompanying drawings.

Brief Description of the Drawings

An embodiment of the invention is now described by way of example and with reference to the accompanying drawings in which: Figure 1 is a block diagram of an activity monitoring system embodying one aspect of the invention; Figure 2 is a schematic diagram of an environment in which part of then activity monitoring system of Figure 1 may be installed; Figure 3 is a table showing some examples of signs that may be detected or determined to be 5 absent by the system of Figure 1, and examples of interpretations of those signs, in particular illustrating how a sequence of more than one events, or absence of events. within a defined period of time may be used to infer a level of "well-being".

Figure 4 is a table showing examples of sensor types that may be included in the system of Figure 1, 10 and possible associated well-being confidence levels; and Figure 5 is a graph illustrating how an intervention parameter may change over time. in particular illustrating how a steady decrease in such a parameter over fime, when no sensory evidence of wellbeing is received, can result in threshold related alerts.

Detailed Description of the Drawings

Referring now to Figures 1 and 2 of the drawings there is shown, generally indicated as 10, a monitoring system embodying one aspect of the invention. The system 10 comprises at least one, but typically a plurality of, sensors 12 for monitoring an environment, preferably including a person (not shown) in the environment. Typically, the sensors 12 are configured to monitor at least one, but typically a plurality of, parameters in an environment 14. The parameters may relate not only to the environment 14 itself, but also to the person in the environment. Each parameter typically relates to a detectable action, event, activity, condition or other sign in, of or relating to the environment, which may include detectable physiological attributes of the person On particular one or more vital sign such as heart rate or respiratory rate). The sensors 12 and the data produced by the sensors 12 may therefore be used to detect or monitor one or more signs, where each sign may relate to the environment and/or the person in the environment, and which advantageously may be used to make inferences about the person's well-being. The environment 14 may be the person's domestic home, or a care home, or any other residence or housing for the person(s) being monitored.

In preferred embodiments, the system 12 includes one or more sensor 12 for monitoring any one or more of the following parameters: * Sounds (e.g. bangs, barks, door bells, telephone ring, coughs, breathing, snoring, sneezes, laughter, a yell or other non-verbal human utterance, kettle boiling or other routine domestic sounds) * Speech * Light * Movement * Heart rate * Respiration rate * Temperature * Use of equipment or objects, e.g. television, cooker. a door or a toilet.

Each sensor 12 may be of any suitable conventional sensor type. Typical embodiments of the system 10 include one or more instance of any one or more of the following sensor types: * Audio sensor (e.g. a microphone, microphone array or sound detector) * Electro-optical sensor (e.g. light sensor, IR sensor) * Image sensor (e.g. a camera (video or stills)) * Motion sensor (e.g. a PIR detector, accelerometer, vibration sensor, piezoelectric movement sensor, pressure sensor or video camera) * Heart-rate sensor (e.g. optical heart rate sensor or pulse sensor, or ECG electrodes, or PCG sensor) * Respiration sensor (e.g. acoustic sensor, or pressure sensor, or piezoelectric sensor) * Ambient temperature sensor * Switch sensor (e.g. electrical, electromechanical, optical or magnetic sensors for detecting operation of equipment or objects).

The monitoring performed by any given sensor 12 may depend on the sensor type and/or on the parameter being monitored and may comprise data capture. simple detection (e.g. presence or absence) and/or measuring, as applicable.

Each sensor 12 may be installed in the environment 14, for example mounted on or incorporated into a wall, ceiling or floor, or incorporated into a device (e.g. a remote control, a television, computing device or household appliance (electrical or otherwise)) or object (e.g. a door, or window, or bed or item of furniture, or any fixture or fitting) that is present in the environment; or may be worn or carried by the person being monitored, e.g. in a wristband, a smart phone or a smart watch. In some cases the sensor 12 may be dedicated for use by the system 10. In other cases existing sensors 12 may be used by the system 10 (e.g. in the case of the accelerometer or microphone of a smart phone or smart watch, or the motion detector(s) or door sensor(s) of an intruder system) with suitable hardware and/or software integration.

In the example of Figure 2, the environment 14 includes one or more door sensors 12A for detecting opening or closing of a door 16. A flush sensor 12B (which may for example be an audio sensor or an electro-mechanical sensor incorporated into or located beside a toilet 18) may be provided for detecting toilet flushes. One or more electrical socket sensor 12C may be provided for detecting operation of electrical socket(s). One or more bed sensor 12D may be incorporated into a bed 20 (typically in the mattress) for monitoring pressure, heart rate, movement and/or respiration (for example the bed may include one or more pressure sensor for detecting the presence of the person in bed, and/or one or more ballistocardiography sensor for detecting heart rate or respiration, and/or one or more microphone for detecting breathing or snoring). One or more motion detector 12E may be provided in a room 22 to detect movement in the room 22. One or more light sensor 12F may be provided in the room 22 to detect a room light being switched on or off. One or more audio detector may be provided in the room 22 for detecting sounds and/or speech. A remote control sensor 12H (e.g. an IR detector) may be provided in the room 22 for detecting operation of a remote control device (e.g. for television 24). The remote control sensor 12H is preferably located on or adjacent the television 24. Each sensor 12 may be provided separately, or more than one sensor 12 may be provided in a common device, e.g. a set-top box 26.

The monitoring system 10 also includes an analysing system 30 for receiving and analysing data captured by the sensors 12. In typical embodiments, the analysing system 30 receives sensor data via a telecommunications network 32, which may comprise one or more computer network and/or one or more telephone network, any part of which may be wired or wireless as required. Typically, the analysing system 30 is provided remotely from the monitored environment 14, in which case the telecommunications network 32 may comprise a Wide Area Network (WAN). In alternative embodiments, some or all of the analysing system 30 may be provided locally with the sensors 12, e.g. in the same home.

A local gateway device 34 (conveniently located in the monitored environment 14) may be provided by which one or more of the sensors 12 communicate with the analysing system 30 via the telecommunications network 32. One or more of the sensors 12 may be enabled for communication directly with the analysing system 30 across the network 32. One or more of the sensors 12 may communicate with a sensor data analyser 36 connected to the network 32. The sensor data analyser 36 is configured to analyse sensor data and to determine what activity, action, event, condition or other sign is represented by the sensor data. For example, the sensor data analyser 36 may be configured to support sound recognition and/or speech recognition. This may be achieved using any conventional sound recognition and/or speech recognition techniques. The analyser 36 may be configured to perform sound classification, and/or other classification, on the sensor data, and may use any conventional classification algorithms and techniques, including machine learning algorithms and techniques to this end. Optionally, the analyser 36 may be configured to perform signal processing on received sensor output signals, for example to condition the signal and/or to extract the sensor data as required. The sensor data analyser 36 provides the processed sensor data to the analysing system 30. Preferably, the sensor data analyser 36 assigns a confidence level, e.g. a probability value, to the processed sensor data indicating the confidence with which it has identified the activity, action, event or other sign represented by the unprocessed or raw sensor data. The sensor data analyser may for example be implemented by one or more suitably configured ASIC, FPGA or other integrated circuit, and/or a computing device with suitably programmed microprocessor(s).

The analysing system 30 includes sensor data receiving means 38 for receiving sensor data directly from the sensor 12, or from the sensor gateway 34, or from the sensor data analyser 36, as applicable. The received sensor data may comprise raw or unprocessed sensor data (e.g. if received directly from a sensor 12 or from the sensor gateway 34), or may comprise processed sensor data comprising an indication of a detected sign, preferably with its assigned confidence level.

Typically, the analysing system 30 includes a sensor data analyser 40 is configured to analyse sensor data and to determine what sign is represented by the sensor data. For example, the sensor data analyser 40 may be configured to support sound recognition and/or speech recognition. This 5 may be achieved using any conventional sound recognition and/or speech recognition techniques.

The analyser 40 may be configured to perform sound classification, or other classification, on the sensor data, and may use any conventional classification algorithms and techniques, including machine learning algorithms and techniques. Optionally, the analyser 40 may be configured to perform signal processing on received sensor output signals, for example to condition the signal and/or to extract the sensor data as required. Preferably, the sensor data analyser 40 assigns a confidence level, e.g. a probability value, to the processed sensor data indicating the confidence with which it has identified the sign represented by the unprocessed or raw sensor data.

The data receiving means 38 is configured to send any received sensor data or output that requires processing to the sensor data analyser 40. The sensor data analyser 40 sends the processed sensor data to data evaluation means 42. Any sensor data received by the data receiving means 38 (e.g. from the sensor data analyser 36 or from any sensor 12 that provides sensor data that is already suitable for evaluation (i.e. preferably being indicative of a detected sign, and preferably with an assigned confidence level) may be send directly to the data evaluation means 42, or otherwise caused to bypass the sensor data analyser 40.

The data evaluaflon means 42 updates one or more parameter depending on the received sensor data. A respective parameter(s) is maintained for each person being monitored by the monitoring system 10. The value of the parameter(s) provides an indication of the well-being of the respective person and so the parameter(s) may be referred to as the intervenflon parameter. One or more action threshold value is set in relation to the parameter value against which the parameter value is compared. When the parameter value reaches (including reaches and passes) any given action threshold value, the system 10 is configured to generate an alert and/or take any other designated action(s). The present embodiment assumes that only one intervention parameter is supported. In alternative embodiments, more than one intervention parameter may be supported. Each parameter may for example relate to different types of detectable sign. The system 10 may take action depending on the value of any one of the intervention parameters, or depending on the respective value of more than one intervention parameters in combination depending on the embodiment.

In preferred embodiments, the data evaluation means 42 is configured to adjust the intervention parameter value in response to determining that the sensor data indicates that one or more relevant sign has occurred or been detected. The adjustment may be instantaneous or may fake place at a predetermined rate, as desired. A relevant sign in this respect may be any sign associated with normal behaviour of the person. For example, relevant signs may include (but are not limited to): 1) Detection of the person's breathing or heartbeat (optionally including a determination that the respiration rate or heart rate is normal) 2) Movement in the environment 3) Use of the TV set or other appliance 4) Making tea or preparing food 5) Using the bathroom 6) Turning a light on or off 7) Non-verbal human utterances, e.g. coughing sniffing, snoring, sneezing, clearing a throat, laughter 8) Opening or closing a door or window 9) Talking, or making or receiving a phone call More generally, in preferred embodiments, a sign may comprise any indication of human behaviour, activity or condition that can be directly or indirectly determined from the sensor data. Such indications can be regarded as signs of life in the environment.

In preferred embodiments, the interventional parameter value is periodically adjusted incrementally so that, over time and in the absence of one or more signs that the person is well being detected, it reaches the action threshold(s), in response to which the system 10 is configured to take one or more interventional action(s).

Preferably, the value of the intervention parameter is adjusted in response to detection of each relevant sign, preferably as each relevant sign is detected, i.e. as the relevant signs occur over time.

Typically, in response to detecting a sign that the person is well, the intervention parameter value is adjusted to defer the interventional action. This may involve adjusting the intervention parameter value away from the action threshold value(s), or resetting the intervention parameter value to an initial value (or other nominal value). If the intervention parameter value is already at the initial value (or limit value) then it may be maintained at that value.

Preferably, in response to detecting a sign that the person is not well, the intervention parameter 30 value is adjusted to expedite the interventional action. This may involve adjusting the intervention parameter value towards the action threshold value(s), or setting the intervention parameter value to one or other of the threshold values.

The intervention parameter may be a well-being parameter in that its value is an indication of a level of confidence that the person is well. As such its value may be maintained, increased or reset to an initial value in response to detecting a sign that the person is well. The parameter value may be decreased over time or in response to detecting a sign that the person is not well, or may be set to one or other of the threshold values in response to detecting a sign that the person is not well.

The intervention parameter may be a risk parameter in that its value is in indication of a level of risk that intervention is required. As such its value may be maintained, reduced or reset to an initial value in response to detecting a sign that the person is well. In this case, the parameter value may be increased over time or in response to detecting a sign that the person is not well. or may be set to one or other of the threshold values in response to detecting a sign that the person is not well.

The intervention parameter may be implemented by a timer (not shown). The timer value may be an indication of a level of confidence that the person is well. As such the timer may be a countdown timer having a value that decreases over time. The counter value may be maintained, increased or reset in response to detecting a sign that the person is well, and may be decreased in response to detecting a sign that the person is not well. Alternatively, the timer value may be an indication of a level of risk that intervention is required. As such the timer may be a count-up timer having a value that increases over time. The counter value may be maintained, decreased or reset in response to detecting a sign that the person is well, and may be increased in response to detecting a sign that the person is not well.

Reference data representing the relevant detectable signs that warrant adjustment of the intervention parameter may be stored by the system 10 in any convenient manner: or may otherwise be made available to the system 10, in particular the analysing system 30. This data may be pre-determined and provided to the system 10 in any conventional manner, and/or may be used, created and/or modified by the system 10 using any convenient convention machine learning (ML) techniques (e.g. during a training period and/or during use of the system 10). The reference data may include an indication of whether each sign is indicative of the person being well (and optionally a level of confidence with which the sign indicates that the person is well) or not well (and optionally a level of confidence with which the sign indicates that the person is not well). Such indications allow the system to determine how to adjust the intervention parameter(s).

Preferably, the relevant signs comprise signs that are indicative of normal behaviour in the environment 14 (and which may therefore be deemed to indicate that the person is well). Reference data representing detectable signs that are deemed to be indicative of normal behaviour may be stored by the system 10 in any convenient manner, or may otherwise be made available to the system 10, in particular the analysing system 30. This data may be pre-determined and provided to the system 10 in any conventional manner, and/or may be created and/or modified by the system 10 using any convenient convention machine learning (ML) techniques (e.g. during a training period and/or during use of the system 10).

The system 10, in particular the data evaluation means 42, sensor data analyser 40 and/or the sensor data analyser 36 as applicable. may use the sign reference data when determining which signs are represented by the sensor data. To this end, the system 10 may be configured to compare the sensor data with the available sign reference data to determine which of the relevant signs, if any, have been detected by the sensors 12. This may be achieved using any conventional data comparison means, including machine learning (ML) whereby, for example the relevant part(s) of the system 10 are trained on the reference data using ML and, once trained, can recognise the relevant signs in the sensor data Some signs may be detected with certainty or virtual certainly (e.g. signs detected by the operation 5 of switch type or ON/OFF type sensors) whereas other signs may be detected with less certainty (e.g. signs that are identified by sound classification). It is preferred to associate each detected sign with an indication of the level of confidence with which it has been identified.

Some signs may be considered to be a more reliable indication of normal behaviour or well-being 10 than others. It is preferred to associate each detected sign with an indication of extent to which (e.g. associated confidence level) the sign is considered to be a reliable indication of normal behaviour, and therefore the well-being, of the person.

Some signs may be detectable directly from the output of one or more of the sensors 12. For example, the output of a microphone may be used to directly detect the presence of speech or other sound, or the output of a light sensor may be used to directly detect that a light has been switched on, or the output of a switch sensor may be used to directly detect that an appliance has been operated, or the output of a motion sensor may be used to directly detect motion. The system 10 may be configured to infer, or indirectly detect, other signs or an indication of well-being from the output of more than one sensor 12 (optionally including sensors of different types) and/or from a combination or sequence of other detected signs, and/or from a combination or sequence of one or more detected sign with the absence of one or more other sign. A combination of detected events, a sequence of events, or "missing" events may be a source of additional information about the current status or well-being of a person being monitored.

By way of example, Figure 3 shows how sequences of two or more events, or detected signs, can be used to infer another sign and/or to infer an indication of well-being. For example, from detection of a doorbell ringing followed by motion being detected, especially in the vicinity of the door, it can be inferred that the person has answered the door, and this is interpreted with certainty or at least a high degree of certainty as a valid sign for indicating normal behaviour (and the intervention parameter may be adjusted accordingly). In contrast, detection of a doorbell ringing followed by no motion being detected is an indication that the person has not answered the door, which may be interpreted as a contra-indication of a sign of life or normal behaviour (and the intervention parameter may be adjusted accordingly). If the doorbell sounds multiple times without subsequent motion being detected this greatly decreases the confidence in the well-being of the person and hence intervention parameter is adjusted by an appropriately large amount. In other examples, from the detection of a fall (e.g. using a vibration sensor or a worn motion detector or camera) followed by detection of one or more other signs such as subsequent motion, raised heart rate, no change in heart rate, a cry or yell, the severity of the fall can be inferred (and the intervention parameter may be adjusted accordingly). Also, if the system 10 renders a reminder to the person, and the person responds, this can be interpreted with certainty or at least a high degree of certainty as a valid sign for indicating normal behaviour (and the intervention parameter may be adjusted accordingly).

In preferred embodiments. the amount by which. or manner in which, the intervention parameter is 5 adjusted may depend on any one or more of: * The type of detected sign. For example a detected sound may be considered to be a poorer indication of well-being than detecting that a TV remote control has been operated or movement has been detected.

* The extent to which (e.g. associated confidence level) the sign is considered to be a reliable indication of the well-being of the person. For example, detection of a healthy heartbeat may be considered a better indication of well-being than detection of movement per se (since movement is possible even when the person is unwell or injured) * The confidence level with which the sign has been identified. For example, detecting that a door was opened with 100% (or almost 100%) confidence may be a better indication of well-being than detecting laughter with a relatively low confidence.

* The time at which the sign took place. For example, detecting the presence of the person in bed (e.g. by a heart rate sensor or respiration sensor) at night may be a better indication of well-being than detecting the person in bed during the day.

In preferred embodiments. the data evaluation means 42 is configured to adjust the intervention parameter by a relatively high amount in response to detecting signs that are deemed to be a relatively reliable indication of normal behaviour or well-being, and by a lower amount in response to detection of signs that are deemed to be a relatively unreliable indication of normal behaviour or wellbeing.

By way of example, Figure 4 shows examples of sensors and confidence levels with which the corresponding sign may be considered to be a reliable indication of the well-being of the person. It will be seen that mattress heart rate sensors, mattress movement sensors, use of a remote control device (both directly at the remote control device and remotely by an IR sensor) and a worn heart rate sensor or motion sensor are all associated with the highest confidence level, while a mattress pressure sensor is associated with a lower confidence level. Detecting motion using a motion sensor can be associated with the highest confidence level. However, the validity of motion detection can be compromised by the presence of a pet in the environment. Optionally, therefore detection of signs using motion sensors may be disabled if pets are present. The level of confidence assigned to signs that are determined using sound classification is variable and may depend on the level of confidence with which the sign was identified by the relevant classification process. The level of confidence assigned to signs that are determined using light classification is variable and may depend on the level of confidence with which the sign was identified by the relevant classification process. Optionally, the system 10 may be configured to disable its audio sensor(s) or to ignore data emanating from the audio sensor(s) if the system 10 detects that a television or other noisy appliance is operating in the environment.

Optionally, a limit value is set for the intervention parameter such that it cannot be adjusted to a value that is further away from the threshold value(s) than the limit value, i.e. decreased above the limit, or decreased below the limit as applicable. The intervention parameter value is preferably initialised to the limit value, and may be reset to the limit, or initial, value upon detection of one or more sign of well-being.

By way of example, the data evaluation means 42 may adjust the intervention parameter value towards the threshold value(s) over time, in particular before detection of the first detected relevant sign, after the last detected relevant sign: and/or between successive detections of relevant signs. The adjustment towards threshold value may begin after each time the parameter value has been adjusted away from the threshold value. As a result, unless sufficient valid signs are detected over time to keep the intervention parameter from reaching a given threshold value: the threshold value will be reached. The system 10 may be said to cause the intervention parameter value to decay over time towards the, or each, threshold value. Adjustments away from the threshold value(s) counteract this decay. Adjustments away from the threshold value(s) are therefore counter-adjustments, i.e. an adjustment in the opposite sense or direction, in comparison with the adjustment towards the threshold value(s) (e.g. if one adjustment increases the parameter value the other adjustment decreases it, and vice versa). The rate at which the intervention parameter value is adjusted towards the threshold value(s) may be selected to suit the environment 14 and/or the person being monitored. For example, a fixed preset rate may be used, or the rate can be calculated to cause the relevant threshold value to be reached when a predetermined period of time has elapsed since, say, the last detected sign.

The data evaluation means 42 may be configured to adjust the intervention parameter value towards the threshold value(s) in response to detection of signs of concern by the sensors 12, e.g. signs that represent abnormal behaviour. This adjustment is a counter-adjustment, i.e. an adjustment in the opposite sense or direction, in comparison with the adjustment performed in response to detecting signs that are indicative of normal behaviour (e.g. if one adjustment increases the parameter value the other adjustment decreases it, and vice versa). The adjustment may be instantaneous or may take place at a predetermined rate, as desired. Signs of concern may for example include any one or more of a person falling, a breakage, a yell, a cry for help, excessive coughing, an unhealthy heart rate or respiration rate, being in bed at an abnormal time, and so on. The amount by which the intervention parameter value is adjusted towards the threshold value(s) in response to detection of such events may depend on the same or similar factors to those outlined above for adjusting the parameter away from the threshold value. As a result, unless sufficient valid signs are detected over time to keep the intervention parameter from reaching a given threshold value, the threshold value will be reached in response to detection of one or more signs of concern.

The analysing system 30 is configured to generate an alert and/or take any other designated action(s) in response to the intervention parameter reaching a given threshold. In the illustrated embodiment, the analysing system 30 includes alert generating means 44 for generating one or more alerts. The alert(s) may take any desired form and may for example comprise any one or more of the following: * An audio or visual alert rendered at the computer or other device of a carer (which device may be part of the analysing system or may be connected to the analysing system 30 via the network 32 or other communications network) * A message, for example an SMS message, MMS message. pager notification. email, voicemail, automated call, in app message or other electronic message, sent to one or more telephone or computing device of one or more carers * A message, for example an SMS message, MMS message, pager notification, automated call, in-app message or other electronic message, sent to a telephone, computing device or other electronic device of the person being monitored (and preferably being located in the monitored environment 14). Preferably such a message would prompt the user to provide a response to the message. If a response is not received by the system 10 then an alert is sent to one or more carers.

The alerts may be generated and sent by any conventional means, typically via the network 32 or other communications network, and the analysing system 10 may be configured to support the generating, sending and receiving of such messages in conventional manner.

Advantageously the system may first prompt the person being monitored to confirm (or otherwise) that they are well e.g. through a set-top box or other device which can respond to speech or a button press in response to the prompt. A positive response may reset the well-being parameter to a high (or highest) value, a negative or non-response would cause the system to progress with alerting a carer.

By way of example, Figure 5 is a graph showing how the intervention parameter value may change over time in response to the detection of valid signs followed by a quiescent period in which no valid sign is detected. In this example, a first and second threshold values TV1, TV2 are set and the arrangement is such that the respective alerts are generated in response to the intervention parameter value falling to the respective threshold value. Therefore the intervention parameter value is relatively high when the person is deemed to be well and falls as the person's well-being is deemed to diminish. In alternative embodiments the opposite arrangement may be adopted whereby the intervention parameter value is relatively low when the person is deemed to be well and rises as the person's well-being is deemed to diminish. In this example, a limit (in this case an upper limit) is set for the parameter value. Initially, the person is in bed as may be detected by a heart rate sensor installed on or in the mattress. Detection of a heathy heart rate is deemed to be a 100% certain indication of the person's well-being and so the parameter value is at its maximum value limit (10 in this example). When the heart rate is no longer detected (which happens at 10:10 in this example and normally corresponds with the person getting out of bed (which may be confirmed by an appropriately located motion detector), the system 10 begins to reduce the parameter value,. The rate of fall is set by the system 10 and in this example is 1% of the parameter value for each minute without detection of valid sign. Subsequently, a motion sensor is activated (at 10:20) and is determined to represent human motion with a probability of 100%. This counts as a valid detected sign and so the system 10 adjusts the parameter value away from the threshold TV1, i.e. increases it in this example, to a value of 9 out of a possible 10. Subsequently, no valid sign is detected for a period of 20 minutes. During this time the system 10 reduces the parameter value at the predetermined rate. At 10:45 the mattress heart rate sensor detects a healthy heartbeat and so the parameter value is raised again. At 11:00 the heart rate is no longer detected (this would normally be the result of the person getting out of bed) and the parameter value falls at the pre-determined rate.

Subsequently, at 11:15, a sound sensor detects a cough with a probability of 90% certainty. This counts as a valid detected sign and so the system 10 adjusts the parameter value to 8.5. Subsequently, no valid sign is detected for a period of 40 minutes during which time the parameter value falls at the predetermined rate. After 20 minutes, the parameter value reaches the first threshold value TV1, which in this example is a warning threshold. In response to the warning threshold TV1 being reached, the system 10 generates one or more warning alerts. For the next 20 minutes the parameter value continues to fall until it reaches the second threshold value TV2, which is an alarm threshold in this example. In response to the alarm threshold TV2 being reached, the system 10 generates one or more alarm alerts.

The analysing system 30 may be implemented by a single computer, or by multiple co-operating processing systems and/or devices as is convenient. The analysing system 30 may include a controller (not shown) for controlling the overall operation of the system. The controller may comprise any suitably configured or programmed processor(s), for example a microprocessor, microcontroller or multi-core processor. The analysing system 30 may implemented using a single processor, multiple processors or a multi-core processor, running a plurality of processes, one of which may be designated as the controller and the others performing the functionality of described herein, including the functionality of any one or more of the data reception means 38, sensor data analyser 40, the data evaluation means 42, the alert generation means 44 and the dashboard 46 as required. Each process may be performed in software, hardware or a combination of software as is convenient. One or more hardware digital signal processors may be provided to perform one or more of the processes as is convenient and as applicable.

Optionally, the system 10 may be configured to cause an in-room (e.g. via audio and the TV) prompt (e.g. Is everything ok?") in the event that no mattress or motion sensor data has been received and a parameter threshold (e.g. a warning threshold) has been reached. The user may use the remote control or voice to confirm. Detecting motion at this point may be recorded as a sign-of-life but without a response to the Is everything ok?" prompt it would not be regarded as reassuring, and would tend to indicate that interventional action is required.

Optionally, the system 10 is configurable such that all or part of its functionality is only available at certain times of the day or night. During the night, sensors 12 installed in the bed, and in particular the mattress: may provide most or all of the sensor data for the system 10. Sensor data from a mattress sensor provides excellent "signs of life" indications during the night, but once the time of day has been reached when the person would normally be expected to have risen (e.g. based on previous detected or learned behaviour(s)), then detecting that they remain in bed may be a sign-of-life, but it's not a sign of well-being and, if detected, may alternatively be handled as an sign of concern.

Advantageously at least one motion detector 12 is provided in the vicinity of the person's bed and configured to detect the person getting out of bed. In cases where data being received from a bed or mattress sensor, e.g. a heart rate sensor, respiration sensor or movement sensor, stops, the motion detector 12 allows the system 10 to distinguish between a cessation of heart rate or respiration or movement data because the person has got out of bed, and a cessation of heart rate, respiration or movement data because the person has died in bed.

The system 10, in particular the analysing system 30, preferably supports the provision of a graphical user interface, which may be referred to as a dashboard, for displaying information relating to the operation of the system 10 and for allowing a user to set any adjustable parameters supported by the system 10. For example, the dashboard may show the current status of the or each resident in a facility in terms of intervention parameter value, location (in-bed, out-of-bed) and open alerts.

The threshold value(s) for generating alerts may be adjusted, conveniently via the dashboard, for example according to a current level of concern for the respective person being monitored, e.g. based on their current health and service level. A person who has recently been ill might have higher threshold value(s), or a faster parameter value decay rate, than normal in order to trigger intervention more quickly. A person paying for a "premium" monitoring service might be configured likewise to receive earlier attention.

Optionally, the system 10 may be configured to support a remote video check, which may precede a physical visit if the resident's privacy levels have been set to allow "remote viewing-when an alert 30 has been raised.

On receipt of an alert (e.g. via in app, sms, email, phone (text to speech) or other electronic message), or in response to an observed declined in a person's intervention parameter value, a carer may have any one or more of the following options' 1. Initiate an audio/visual prompt from the gateway 34 or other suitable device connected to the system 10 to confirm that everything is ok 2. Listen in via the gateway 34 (eavesdrop) or other suitable device connected to the system 10 for sounds of distress 3. Via a still image from a camera included in the system 10 (e.g. captured once per minute say) 4. View live video stream a video camera included in the system 10 5. View an image or feed from a thermal camera image included in the system 10.

6. Make an automatically answered video call to the person 7. Make a normal (answer required) video call to the person In preferred embodiments, the dashboard communicates at least the intervention parameter value outset but in time should display at least some and preferable all of the following 5 insights into wellbeing: 1. Level of "confidence" in whether person is alive (e.g. the intervention parameter value) 2 Last known sign 3. Environment: is the environment safe? (e.g. are parameters such as temperature, humidity and/or lighting appropriate for the time of day?) 4. Physical well-being: are the person's life signs as expected (within their normal ranges) and are they physically behaving as expected? 5. Mental well-being: is there evidence of sufficient social engagement, healthy pursuit of interests and personal care? A high level dashboard should give immediate "at a glance" status of the person or persons within a carer's domain. The main focus should be on signs-of-life and last detected sign but with environmental, physical and mental well-being preferably being indicated more subtly.

The dashboard preferably highlights or prioritizes those people who are most in need of attention. The system 10, and in particular the dashboard, preferably supports creation of views which correlate to the physical layout of the building e.g. view only those people on a given floor.

When the status has been acknowledged e.g. an alert has issued, it should be clear that the alert has been raised and whether or not it has been acknowledged by anyone.

Acknowledging an alert preferably does not hide it from view in the dashboard, it should instead show that it is being investigated. An alert is preferably explicitly closed (with an appropriate reason entered by the user) before monitoring resumes. When closing an alert the carer should choose whether to reset the intervention parameter value to maximum (e.g. after a physical visit) or not. Preferably, the system 10 supports suspension of monitoring of an individual (they may have died and further alerts are not helpful).

The invention is not limited to the embodiment(s) described herein but can be amended or modified without departing from the scope of the present invention.

Claims

CLAIMS: 1. A method of monitoring an environment using at least one sensor, the method comprising: monitoring said environment using said at least one sensor and generating corresponding sensor data; analysing said sensor data to detect at least one sign that is indicative of life in said environment; providing at least one intervention parameter for a person in said environment; taking at least one interventional action depending on a value of said at least one intervention parameter; incrementally adjusting the value of said at least one intervention parameter towards at least one threshold value that causes said at least one interventional action to be taken; and counter-adjusting the value of said intervention parameter to defer said at least one interventional action in response to detection of at least one relevant sign.
2. The method of claim 1, wherein said monitoring involves monitoring at least one parameter of said environment using said at least one sensor and generating corresponding sensor data.
3. The method of claim 1 or 2, including comparing the value of said intervention parameter to at 20 least one action threshold value, and taking said at least one action depending on the comparison.
4. The method of any preceding claim, wherein said counter-adjusting involves counter-adjusting the value of said intervention parameter away from said at least one threshold value or resetting the intervention parameter value to a limit value.
5. The method of any preceding claim, wherein taking at least one action involves generating at least one alert.
6. The method of claim 5, wherein generating said at least one alert involves sending at least one 30 electronic message and/or rendering at least one visual and/or audio signal and/or initiating a telephone or video call.
7. The method of any preceding claim, wherein said at least one relevant sign comprises at least one sign that is indicative of normal behaviour in said environment and/or of a normal condition of a 35 human in said environment.
8. The method of any preceding claim, wherein said counter-adjusting involves adjusting the value of said intervention parameter in response to detection of each relevant sign, preferably as each relevant sign is detected.
9. The method of any preceding claim, wherein said counter-adjusting involves adjusting the value of said intervention parameter by an amount or in a manner that is determined by a type of said at least one relevant sign.
10. The method of any preceding claim, wherein said counter-adjusting involves adjusting the value of said intervention parameter by an amount or in a manner that is determined by a confidence level with which said at least one relevant sign is detected.
11. The method of any preceding claim, wherein said counter-adjusting involves adjusting the value 10 of said intervention parameter by an amount or in a manner that is determined by a confidence level with which said at least one relevant sign is deemed to indicate the person's well-being.
12. The method of any preceding claim, wherein said intervention parameter has a limit value. said intervention parameter value preferably being initialised to said limit value.
13. The method of any preceding claim, wherein said adjusting involves adjusting the value of said intervention parameter at an adjustment rate, wherein said adjustment rate may be fixed, or may be calculated based on a desired time taken for said counter-adjusting to adjust said intervention parameter to reach said at least one threshold value.
14. The method of any preceding claim, further including adjusting the value of said intervention parameter value to expedite taking said at least one interventional action in response to detecting at least one abnormal sign that is indicative of abnormal behaviour and/or an abnormal human condition, and/or in response to detecting an absence of said at least one relevant sign, wherein adjusting the value of said intervention parameter value to expedite taking said at least one interventional action may involve adjusting the value of the intervention parameter value towards said at least one threshold value, or setting the value of the intervention parameter to said at least one threshold value.
15. The method of claim 14, wherein said adjusting involves the value of said intervention parameter in response to detection of each abnormal sign, preferably as each abnormal sign is detected.
16. The method of claim 14 or 15, wherein said adjusting involves adjusting the value of said intervention parameter by an amount or in a manner that is determined by a type of said at least one 35 abnormal sign.
17. The method of any one of claims 14 to 16, wherein said adjusting involves adjusting the value of said intervention parameter by an amount or in a manner that is determined by a confidence level with which said at least one abnormal sign is detected.
18. The method of any one of claims 14 to 17, wherein said counter-adjusting involves adjusting the value of said intervention parameter by an amount or in a manner that is determined by a confidence level with which said at least one abnormal sign is deemed to indicate the person's well-being.
19. The method of any preceding claim, wherein said at least one sensor comprises one or more instances of any one or more of the following sensors: audio sensor; electro-optical sensor; image sensor; motion sensor; heart rate sensor; respiration sensor; ambient temperature sensor; pressure sensor and/or switch sensor.
20. The method of claim 2, wherein said at least one parameter of the environment comprises any one or more of the following parameters: sounds; speech; light; movement; heart rate; respiration rate; temperature; use of equipment or objects.
21. The method of any preceding claim, wherein said at least one sign may include any one or more of: detection of the person's respiration or heart rate (optionally including a determination that the respiration rate or heart rate is normal); movement in the environment; use of an appliance; preparation of food or drink; using the bathroom; turning a light on or off; non-verbal human utterances; opening or closing a door or window; talking, or making or receiving a phone call.
22. The method of any preceding claim, wherein said at least one intervention parameter is implemented by at least one timer.
23. The method of any preceding claim, wherein said analysing involves comparing said sensor data with reference data representing detectable signs.
24. The method of any preceding claim, wherein said analysing involves classifying said sensor data to detect said at least one sign.
25. A monitoring system for monitoring an environment, the system comprising; at least one sensor for monitoring said environment and generating corresponding sensor data; analysing means for analysing said sensor data to detect at least one sign that is indicative of life in said environment; means for providing at least one intervention parameter for a person in said environment; means for taking at least one interventional action depending on a value of said at least one intervention parameter; means for incrementally adjusting the value of said at least one intervention parameter towards at least one threshold value that causes said at least one interventional action to be taken; and means for counter-adjusting the value of said intervention parameter to defer said at least one interventional action in response to detection of at least one relevant sign.