False data reduction in tracking a human target in an indoor environment

Abstract

There is provided an apparatus, a system, a computer-implemented method, and computer software for tracking one or more human targets in an indoor environment. The method includes receiving sensor data indicative of a location, at a plurality of timepoints, of a first sensed target. The sensor data is obtained by a sensing means and includes coordinate information of the first sensed target at each timepoint of the plurality of timepoints. The method further includes analysing a motion of the first sensed target based on the received sensor data. The method also includes classifying the first sensed target as a human target or a false target based on a result of the analysis of the motion.

Description

FALSE DATA REDUCTION IN TRACKING A HUMAN TARGET IN AN INDOOR ENVIRONMENT

[0001] This invention relates to methods and systems for tracking a human target in an indoor environment. In particular, the invention relates to monitoring a location of a vulnerable person in a room within a secure care setting, and to distinguishing between genuine signals and false signals.BACKGROUND

[0002] Monitoring the safety of a vulnerable person poses many practical and ethical challenges. For example, settings such as mental health institutes, prisons, addiction recovery centres and care homes may contain service-users who are at risk of causing serious physical harm to themselves or others. These service-users require constant monitoring to reduce this risk and ensure that instances of personal harm can be detected and dealt with quickly. The service-user may be monitored using optical cameras or physical observations by staff. However, physical observations place a high burden on staff time. Moreover, both optical cameras and physical observations can feel intrusive to the service-user, especially when the service-user is undertaking intimate personal care activities.

[0003] Non-visual sensors, such as radar sensors, allow the service-user to be monitored whilst respecting their privacy. However, these sensors are susceptible to errors such as false positive readings, false negative readings, and inaccurate location information of a service-user. Sensing errors can lead to problems such as incorrectly identifying whether a service-user requires assistance and undue staff burden.

[0004] It is in this context that this invention is devised.BRIEF SUMMARY OF THE DISCLOSURE

[0005] In accordance with the present invention, there is provided an apparatus, a system, a computer-implemented method, and computer software for tracking one or more human targets in an indoor environment.

[0006] According to a first aspect, there is provided a computer-implemented method for tracking one or more human targets in an indoor environment. The method includes receiving sensor data indicative of a location, at a plurality of timepoints, of a first sensed target. The sensor data is obtained by a sensing means and includes coordinate information of the first sensed target at each timepoint of the plurality of timepoints. The method further includesanalysing a motion of the first sensed target based on the received sensor data. The method also includes classifying the first sensed target as a human target or a false target based on a result of the analysis of the motion.

[0007] Advantageously, this method enables false targets to be identified and dealt with accordingly. False positive readings incorrectly indicate to staff that there are additional people present in a room. Having more than a set number of people in a room may be a safety or security risk which requires staff to take action. By identifying false positive readings, these readings may then be disregarded or otherwise mitigated so that staff attention is not unnecessarily directed to false alarms. False negative readings incorrectly indicate to staff that a person is not present in a room. There is then a risk that a safety of the person is compromised and staff are not aware so are unable to help. By identifying false negative readings, staff may be alerted to perform a manual check on the wellbeing of the person in the room, thereby reducing a likelihood of the person being harmed.

[0008] When a sensing means receives a plurality of return signals from a single target, where the return signals have taken different paths to the sensing means, the return signals may overlap reduce the resolution of the sensing means. This method further advantageously, improves the spatial resolution of the sensing means by distinguishing between reflected and true signals. This can allow a location of a person in the room to be more accurately identified. A vulnerable person may be more likely to come to harm in some regions of an indoor environment than in other regions of the indoor environment. For example, doorways are known ligature risk areas. By improving the accuracy of the detected location of the person, it can be more reliably determined whether the person is in a high risk area of the indoor environment. Prompt action can then be taken by staff if the amount of time spent that a detected person spends in a high risk area is a cause for concern.

[0009] Optionally, analysing the motion of the first sensed target includes extracting, from the received sensor data, one or more motion features indicative of a characteristic of a movement of the first sensed target. Analysing the motion of the first sensed target further includes comparing the extracted one or more features to a predetermined model of expected motion patterns of one or more sensed target classifications in the indoor environment.

[0010] Optionally, the extracted one or more motion features includes an initial location of the first sensed target where the initial location is within the indoor environment. Analysing the motion of the first sensed target further includes comparing the initial location of the first sensed target to a predetermined location of one or more entryways into the indoor environment, and classifying the first sensed target as the human target or the false target based on a result of the comparison.

[0011] Advantageously, this method identifies targets which have materialized in an unusual or unexpected place which is inconsistent with the behaviour of a human. Such targets can then be readily designated as false targets which are likely due to environmental factors or inanimate objects.

[0012] Optionally, the extracted one or more motion features comprises a range of movement of the first sensed target. Analysing the motion of the first sensed target further comprises comparing the range of movement of the first sensed target to a predetermined threshold value and classifying the first sensed target as the human target or the false target based on a result of the comparison.

[0013] Advanageously, this enables a non-visual sensor to distinguish between a human and an inanimate object because a human is likely to have a greater range of movement than an inanimate object.

[0014] Optionally, the computer-implemented method further includes receiving sensor data indicative of a location of a second sensed target at the plurality of timepoints. The extracted one or more motion features comprises a velocity coefficient of the first sensed target. The velocity coefficient relates to a ratio of a velocity of the first sensed target compared to a velocity of the second sensed target. Classifying the first sensed target as the human target or the false target comprises comparing the velocity coefficient of the first sensed target with a second predetermined threshold value and classifying the first sensed target based on a result of the velocity coefficient comparison.

[0015] Advantageously, this enables targets which result from a reflection of a human target to be identified and subsequently classified as false targets. Reflective surfaces, such as mirrors, can give the illusion that there are two targets in an indoor environment. By identifying reflected targets as false targets, the reflected targets may be disregarded. This reduces the likelihood that an alarm indicating a presence of two or more people in an indoor environment will be incorrectly output.

[0016] Optionally, the computer-implemented invention comprises receiving sensor data indicative of a location of a third sensed target at the plurality of timepoints. The extracted one or more motion features comprises a distance of the first sensed target from the sensing means. Classifying the first sensed target as the human target or the false target comprises comparing the distance of the first sensed target with a distance of the third sensed target from the sensing means and classifying the first sensed target based on a result of the distance comparison.

[0017] Optionally, the computer-implemented invention comprises assigning a continuous target identifier, ID, to each sensor reading corresponding to the first sensed target.

[0018] Advantageously, assigning a continuous target ID to a sensed target enables the target to be tracked more reliably.

[0019] Optionally, assigning the continuous target ID comprises predicting, for time t, a location of the first sensed target at time t+1 based on the sensed location of the first target at time t; calculating a distance between a sensed reading at t+1 and the predicted location; assigning an existing ID to the first sensed target at t+1 if the calculated distance is less than a predetermined threshold value; and assigning a new ID to the first sensed target at t+1 if the calculated distance is greater than the predetermined threshold value.

[0020] Optionally, in a case that there are a plurality of sensed readings at t, and a plurality of sensed readings at t+1 , assigning the continuous target ID further comprises determining a distance between each predicted location and each sensed reading at t+1 ; and assigning each predicted location to one of the plurality of sensed readings at t+1 by minimising a cost function based on the determined distances between the predicted locations and the sensed readings and a threshold penalty.

[0021] Advantageously, this enables a continuous ID to be assigned to each of the plurality of sensed targets so that each target may be more reliably tracked.

[0022] Optionally, Unscented Kalman Filtering, UKF, is used to predict the location of the target at t+1.

[0023] Advantageously, UKFs perform well with non-linear data, and human behaviour is inherently non-linear.

[0024] Optionally, a Kuhn-Munkres algorithm is used to assign the continuous target ID to the first sensed target.

[0025] Optionally, the computer-implemented method also includes determining whether the first sensed target has exited the indoor environment.

[0026] Advantageously, this enables a non-visual sensing means to distinguish between a human who appears to have vanished and a human who has left the room. When a person is unable to be sensed by the non-visual sensing means, this can be a cause for concern because it may indicate that the person is in a harmful situation. However, if the person is unable to be sensed because they have left the room, this is not necessarily indicative of the person potentially being in an unsafe situation. Being able to distinguish between a person who has left the room and a person who is still in the room but not detectable by the sensing means reduces the risk of false negative sensor readings. It also provides a more reliable indication of when a person requires assistance, so prevents staff members from being fatigued by false alarms.

[0027] Optionally, determining whether the first sensed target has exited the indoor environment comprises determining an exit location of the first sensed target. The exit location is within the indoor environment and comprises coordinate information of the first sensed target obtained at a last time point of the plurality of time points where the first sensed target is still within the indoor environment. Determining whether the first sensed target has exited the indoor environment also includes comparing the exit location of the first sensed target with a predetermined location of one or more exits out of the indoor environment; and determining that the first sensed target has exited the indoor environment if the exit location of the sensed target is within the predetermined location of one of the one or more exits out of the indoor environment.

[0028] Optionally, the computer-implemented method also includes deleting the sensing data corresponding to the first sensed target from a memory if no further sensor readings of the first sensed target are obtained in the indoor environment for at least a predetermined time period after the first sensed target has exited the indoor environment.

[0029] Advantageously, this prevents memory creep and reduces a storage burden on the memory.

[0030] Optionally, the computer-implemented method also includes deleting the sensing data corresponding to the first sensed target from a memory if the first sensed target is classified as the false target.

[0031] Advantageously, this frees up space within the memory by preventing data relating to false targets from unnecessarily taking up memory space.

[0032] According to another aspect there is provided an apparatus for tracking a human target in an indoor environment. The apparatus includes a communication module. The communication module is configured to receive sensor data indicative of a location of a first sensed target at a plurality of timepoints. The sensor data is obtained by a sensing means and comprises coordinate information of the first sensed target at each timepoint of the plurality of timepoints. The apparatus also includes a memory configured to store the received sensor data. The apparatus also includes a processor. The processor is configured to analyse a motion of the first sensed target, based on the received sensor data; and classify the first sensed target as a human target or a false target based on a result of the analysis of the motion.

[0033] According to yet another aspect of the invention, there is provided a system for tracking a human target in an indoor environment. The system comprises the apparatus for tracking a human target in an indoor environment. The system further includes one or more radio frequency sensors configured to obtain sensor data indicative of a sensed human target in the indoor environment.

[0034] According to a further aspect of the invention, there is provided computer software which, when executed, is arranged to perform a method for tracking a human target in an indoor environment.BRIEF DESCRIPTION OF THE SEVERAL VIEWS OF THE DRAWINGS

[0035] Embodiments of the invention are further described hereinafter with reference to the accompanying drawings, in which:

[0036] FIG. 1 is a block diagram of an apparatus 100 according to an embodiment of the invention.

[0037] FIG. 2 is a schematic of a system 200 according to an embodiment of the invention.

[0038] FIG. 3 shows a flow chart of a method 300 according to an embodiment of the invention.

[0039] FIG. 4 shows a flow chart of a method 400 according to an embodiment of the invention.

[0040] FIG. 5 shows a flow chart of a method 500 according to an embodiment of the invention..

[0041] FIG. 6 shows a flow chart of a method 600 according to an embodiment of the invention.

[0042] FIG. 7 shows a flow chart of a method 700 according to an embodiment of the invention..

[0043] FIG. 8A is a photograph of an indoor environment 800 according to an embodiment of the invention .

[0044] FIG. 8B is an illustration of a first experimental set up.

[0045] FIG. 8C is an illustration of a second experimental set up.

[0046] FIG. 9 shows results obtained by the first experimental set up.

[0047] FIG. 10 shows results obtained by the second experimental set up.DETAILED DESCRIPTION

[0048] FIG. 1 shows an apparatus 100 for identifying a sensed target as a human target or a false target according to an embodiment of the invention. The apparatus 100 may be referred to as, for example, a controller or controlling means. The apparatus 100 comprises a communication module 102, a memory 104, and a processor 106.

[0049] The communication module 102 is configured to receive sensor data from one or more sensing means. The communication module 102 may comprise one or more antennas, receivers, or transceivers configured to receive the sensor data. The sensor data provides information related to a location of a sensed target in an indoor environment, such as coordinate information of the sensed target. The sensor data is obtained at a plurality of timepoints andcoordinate information of a sensed target may be obtained at each timepoint of the plurality of timepoints.

[0050] The indoor environment may be a room, such as a bedroom or a bathroom. The room may be located in a mental health care setting such as a hospital, care home or other institution.

[0051] The sensed target is a target detected by the sensing means. A motion of a sensed target is sensed or detected by the sensing means. The target may be referred to as an object, person, human, artifact or entity. A motion of the sensed target may be tracked or monitored. Tracking or monitoring the motion of the sensed target may comprise obtaining coordinate information of the object at a plurality of time points. Tracking or monitoring the motion of the sensed target may further comprise performing a motion pattern analysis of the target based on the obtained coordinate information.

[0052] The sensed target may be a person such as a staff member or a patient. When the sensed target is a person, the sensed target may be referred to as a human target, true target or real target.

[0053] The sensed target may be an inanimate object sensed by the sensor 202, perhaps as a result of a movement of the inanimate object. For example, a movement of a door or a fan may trigger the sensor 202 and cause the sensor 202 to obtain location information of the door or fan. This is a false positive signal reading, and the sensed target may be referred to as a false target.

[0054] The sensed target may be an artifact caused by the environment within the indoor environment. For example, the sensed target may result from signals reflected or attenuated by moisture or reflective surfaces within the indoor environment. When the reflected or attenuated signals are detected by the sensing means, they can provide the illusion of a person being present in a location when no such person is actually present in that location. This is a false positive signal reading, and the sensed target may be referred to as a false target or reflected target.

[0055] The memory 104 is configured to store the sensor data received by the communication module 102. The memory 104 is further configured to store computer-readable instructions which when executed, cause the processor 106 to perform methods according to the present invention. The memory 104 may also be configured to store one or more outputs generated by the processor 106 as a result of the methods performed by the processor 106.

[0056] The processor 106 is configured to process the sensor data and determine whether the sensed target is a human target or a false target based on a result of the processing. The processor 106 processes the sensor data to analyse a motion of the sensed target. Theprocessor 106 then classifies the sensed target as a human target or a false target based on a result of the analysis of the motion

[0057] Figure 2 shows a system 200 for identifying a sensed target as a human target or a false target in accordance with an embodiment of the invention.

[0058] The system 200 comprises an apparatus 100 and one or more sensing means 202. The one or more sensing means 202 may be referred to as a sensor 202. The sensor 202 obtains real-time location data of one or more people located within a range of the sensor 202. Where the range of the sensor 202 extends beyond the boundaries of the indoor environment, the sensor 202 may selectively obtain real-time location data of one or more people located within the boundaries of the indoor environment.

[0059] The sensor 202 is a non-visual and non-contact sensor for detecting a presence of a person in the indoor environment and obtaining location data of the sensed person. The sensor 202 may be a millimeter wave (mmWave) radio frequency radar.

[0060] By ‘non-contact’, it is intended to mean that, in normal use, the sensor 202 is not in physical contact with a person who is being sensed by the sensor 202. For example, the sensor 202 may be mounted on a ceiling of the indoor environment.

[0061] The sensor 202 is communicatively coupled to the apparatus 100 through the communication module 102. The sensor 202 may be connected to the apparatus 100 via a wired connection, or may be coupled wirelessly. In some embodiments, the sensor 202 may be connected to the apparatus 100 indirectly via one or more networks such as a local area network (LAN) or the Internet. The sensor 202 is configured to transmit location information of one or more sensed targets to the apparatus 100. The location information may comprise x, y, z coordinates of a sensed person at each timepoint of a plurality of timepoints. The sensor 202 is configured to continuously or periodically transmit the location information to the apparatus 100, such that the apparatus 100 is provided with a time series of location information of a sensed target.

[0062] FIG. 3 shows an example flow chart of a method 300 of tracking a human target in an indoor environment. The method 300 may be performed at least in part by an apparatus such as the apparatus 100. In particular, the method 300 may be performed by a processor 106 of the apparatus 100.

[0063] In step 302, sensor data indicative of a location of a sensed target is received. The sensor data includes location information of the sensed target at a plurality of timepoints. The sensor data is obtained by a sensing means, such as sensor 202, and comprises coordinate information of the sensed target at each timepoint of the plurality of timepoints. The sensedtarget is any target, human or false, from which the sensing means obtains sensor data. For example, the sensed target may be a person located within an area of the indoor environment. As another example, the sensed target may be a reflection of a person in the indoor environment, the reflection coming from a mirror or other reflective surface of an object within the indoor environment. As another example, the first sensed target may be an inanimate object located within the indoor environment.

[0064] In step 304, the apparatus 100 analyses a motion of the sensed target based on the received sensor data. Since the received sensor data comprises a series of coordinates of the sensed target at different time points, a movement of the sensed target over time can be determined. By analysing the movement over time, it is possible to determine whether the sensed target is likely to be a human target or a false target. This analysis may be referred to as a motion pattern analysis.

[0065] In step 306, the apparatus 100 classifies the sensed target as a human target or a false target. The classification is performed based on a result of the analysis of the motion. If the motion analysis indicates that the sensed target is more likely to be a human target than a false target, then the sensed target is classified as a human target. If the motion analysis indicates that the sensed target is more likely to be a false target than a human target, then the sensed target is classified as a false target.

[0066] In step 306, the apparatus may further store the sensor data along with the corresponding target classification in a memory. The memory may be a local memory, such as the memory 104, or a memory of another device. The memory may be a cloud storage means. The sensor data and classified status may be saved to a log file in the memory. The log file may comprise a series of timepoints and corresponding coordinate information of one or more sensed targets. The log file may further comprise the classification determined for each of the one or more targets so that the classification is associated with the corresponding target.

[0067] FIG. 4 shows a flow chart of a method 400 for analysing the motion of the sensed target, also known as performing a motion pattern analysis of the sensed target. That is, FIG. 4 shows at least some of the steps which may be occur when performing step 304 of FIG. 3. The method 400 may be performed by the apparatus 100 and, in particular, by the processor 106 of the apparatus 100.

[0068] In step 402, one or more motion features are extracted from the received sensor data. A motion feature relates to a characteristic of the motion of the sensed target. For example, a motion feature may comprise one or more of a starting location, a velocity, an end location, and a distance of the sensed target from the sensor. The starting location and the end location relate to locations within the designated indoor environment and may not, for example, relate to alocation outside of the room (such as in an adjacent room or a corridor), even if the range of the sensor is such that targets outside of the room may be sensed.

[0069] In step 404, the one or more extracted motion features are compared to a predetermined model. The predetermined model is based on motion traits which are unique to human targets. The predetermined model maps the set of assumptions about the motion behaviour of human targets to a set of requirements. The one or more extracted motion features are compared to the set of requirements to determine whether the extracted motion features relate to a characteristic of a movement of a human target, or to a characteristic of a false target.

[0070] As a first example, it is known that human targets enter a room at a valid entry point, such as a doorway. In contrast, false targets may first appear in a random location in the room, which may give the illusion that they have entered through a wall or spontaneously appeared in a center of the room. The terms entry point and entryway may be used interchangeably.

[0071] Based on the assumption that human targets enter a room at a valid entry point, a requirement for a human target may be that the sensor data originates at a valid entry point. In this case, the extracted motion feature may be a location in the room where the sensed target is first sensed. This location may be referred to as an initial location, a first location, a starting location, or an origin of the sensed target. In more detail, the extracted motion feature may be x, y, z coordinates of the sensed target at time t = 0, where time t = 0 is the first time that the sensed target is sensed within the indoor environment for at least a predetermined period of time. Time t = 0 may be referred to as time t, a starting time, a first time, or a time of a first sensor reading of the sensed target within the indoor environment. Of course, a person may leave and re-enter a room multiple times. Each occurrence of a person entering a room may be considered a first time, even when the same person has previously entered the room on one or more other occasions.

[0072] The requirement may be that the initial location of the sensed target within the room is within a predetermined location corresponding to a valid entryway into the room. In other words, a requirement for a human target may be that a first sensor reading of the target in the room is within a predetermined distance of a valid entryway into the room. If the first sensor reading of the target in the room is outside the predetermined distance of the valid entryway, then this is indicative of the target being a false target.

[0073] The predetermined distance may be set based on a sensing frequency of the sensing means. For example, for a first sensing means having a sensing frequency of 20ms and a second sensing means having a sensing frequency of 40ms, the predetermined distance may be smaller for the first sensing means than the second sensing means. For example, the predetermined distance when using the first sensing means may be 0.5m whereas thepredetermined distance when using the second sensing means may be 1 m. These numbers, and the relationships between the numbers, are merely examples and are not intended to be limiting.

[0074] The predetermined distance may be set based on a sensing frequency of the sensing means and a typical walking or running speed of a human. For example, the distance may be set to be a maximum distance expected to be moved by a person during the interval between consecutive sensor readings.

[0075] The predetermined location may be an area of the room between the entryway and a boundary line at the predetermined distance. That is, the predetermined location may be the portion of the room within a certain radius of the entryway, where the radius is the predetermined distance.

[0076] As a second example of one or more extracted motion features being compared to a predetermined model, it is known that false targets created by inanimate objects generally remain within a fixed location. In contrast, a human target generally has a larger range of movement. Based on the assumption that false targets remain in fixed locations, whereas human targets move around the room, a requirement for a human target may be having a range of at least a predetermined area within a set time period. In this case, the extracted motion feature may be a range of the sensed target over a set time period. The requirement may be that the extracted range over the set time period is greater than a predetermined range. The x and y coordinates of the location of the sensed target may be used to determine the range of the sensed target.

[0077] For example, the requirement for a human target may be having a range of at least 1 m2over a past two seconds. If the range of the sensed target is less than the predetermined area within the set time period, this may be indicative of the sensed target being a false target. Of course, a person may also have a range which is less than the predetermined area within the set time period, such as when the person is resting, reading, or otherwise located in a same place for a length of time. The range being less than the predetermined area within the set time period is therefore not a reliable indication of whether the target is human or false. However, when the range is greater than or equal to the predetermined area within the set time period, this may be a good indication that the sensed target is a human target rather than a false target.

[0078] As a third example of one or more extracted motion features being compared to a predetermined model, it is known that reflected targets have a high correlation of velocity with at least one other target. For example, where there are two sensed targets, where one sensed target corresponds to a person and the other sensed target relates to a reflection of the person, a velocity of the two sensed targets is likely to be highly correlated. This is because the motionof the reflected person is likely to be dependent on the motion of the person. In contrast, where the two sensed targets correspond to two different people in the indoor environment, each of the two sensed targets may move independently. Since the motion of the first person is different to, and independent of, the motion of the second person, a velocity of the two sensed targets may not be highly correlated.

[0079] Based on the assumption that a reflected target has a high correlation of velocity with at least one other target, the requirement may be that a correlation coefficient of velocity between a first sensed object and a second sensed object is greater than a predetermined threshold correlation coefficient. The requirement may be that the correlation coefficient of velocity is greater than the predetermined threshold correlation coefficient over a minimum predetermined period of time. For example, the requirement may be that a Pearson correlation coefficient of velocity is greater than 0.8 over a time duration of 2s. When the requirement is met (such that the correlation coefficient is greater than the threshold value), this may indicate that at least one of the first sensed target and the second sensed target is a false target. When the requirement is not met (such that the correlation coefficient of velocity is less than the threshold value), this indicates that the first sensed target and the second sensed target are likely to be human targets. The x, y, and z coordinates of the location of the sensed targets may be used to determine the velocity coefficient.

[0080] As a fourth example of one or more extracted motion features being compared to a predetermined model, it is known that a reflected target should be further from the sensing means than the true target from which the reflected target originates. This is due to Pythagorus' theorem and the laws of reflection. The requirement may be that the distance of a true target from the sensing means is less than the distance of a reflected target from the sensing means.

[0081] Where a first target and a second target are sensed in the indoor environment, a distance of each target from the sensing means may be calculated. The calculated distances may be compared, and it may be determined that the target which is closer to the sensing means is a human target, and the target which is further from the sensing means is a false target. Of course, it may be that both of the targets are human targets, where one person is naturally further from the sensing means than the other person. The furthest sensed target therefore cannot automatically be designated as a false target. As such, the relative distances of the sensed targets from the sensing means is unlikely to be used in isolation but as part of a set of factors for determining whether a sensed target is a true target or a false target.

[0082] The abovementioned assumptions of motion patterns for real and false targets are summarised in the table below.

[0083] Using the assumptions of motion patterns, a status may be assigned to each sensed target. All newly sensed targets may be assigned a status of “ghost target”. That is, all new targets may begin as a ghost target.

[0084] When a ghost target has met the requirements of assumptions 1 and / or 2 in the above table (such that the ghost target originates at a valid entry / exit point and / or has a range of at least a threshold value), the status of the ghost target may be updated to “true target”. If a ghost target instead meets the requirements of assumptions 3 and / or 4 (such that the ghost target has a high correlation of velocity with another sensed target in the indoor environment, and / or the ghost target is further from the sensing means than the another sensed target), the status of the ghost target may be updated to “reflected target” or “false target”.

[0085] These assumptions and requirements are non-limiting examples. Not all assumptions may need to be met in order to classify a target, and other assumptions may be used in addition to (or instead of) any of the aforementioned assumptions.

[0086] FIG. 5 shows an example flow chart of another method 500 for tracking a human target in an indoor environment. Some of the features of method 500 are the same as those of method 300. For brevity, a detailed description of those features will not be repeated. The method 500 may be performed by the apparatus 100, more particularly by the processor 106.

[0087] In step 302, the apparatus receives sensor data of the sensed target. The sensor data is obtained from the sensing means.

[0088] In step 502, a continuous identifier is assigned to one or more datapoints of the received sensor data in order to identify target continuity over time.

[0089] Sensing means can inherently lack the capability to distinguish between targets from one moment to the next. That is, the sensing means may not readily recognise that the different locational data obtained at time t and at time t+1 may relate to a same target. However, it is important to correctly match locational data from successive timepoints to a corresponding target to achieve an accurate motion pattern analysis of the target. By assigning a continuous identifier to datapoints relating to a same target, the target may be more accurately monitored and a location of the target more accurately obtained. Step 502 is described in more detail in relation to FIG. 6.

[0090] In step 304, the motion of the sensed target is analysed. In other words, a motion pattern analysis is performed on the sensed target. The motion may be analysed according to the steps of method 400.

[0091] In step 306, the sensed target is classified as a human target or a false target based on a result of the motion pattern analysis. The sensor data and the result of the classification may be stored in a memory.

[0092] In step 504, it may be determined whether the first sensed target has exited the room. This step is described in more detail in relation to FIG. 7.

[0093] In FIG. 5, step 504 is provided in a dashed border to indicate that this step may be skipped before moving on to step 506. That is, step 506 may not be dependent on step 504 being successfully completed.

[0094] In step 506, sensor data relating to one or more of the sensed targets may be deleted. The data may be deleted from the memory.

[0095] As a first example, sensor data relating to a target which has been classified as a false target may be deleted. In this case, it is not necessary to perform step 504 prior to performing step 506.

[0096] As a second example, where step 504 has been performed and it is determined that the sensed target has exited the room, sensor data relating to the exited target may be deleted. The data (including sensor data and the classification result) relating to the exited target may be deleted after a predetermined amount of time has passed since the target was sensed in the indoor environment. For example, the data relating to the exited target may be deleted after thetarget has left the room and no new sensor readings, or no sensor readings relating to a new sensed target, have been identified for a period of 10s from the target leaving the room.

[0097] FIG. 6 shows an example of a method 600 for assigning a continuous identifier to a sensed target. FIG. 6 shows at least some of the steps which may be performed when carrying out step 502 of FIG. 5. At least some of these steps may be performed by the apparatus 100, in particular by the processor 106.

[0098] In step 602, a prediction is made at each timepoint t for the location of the target at time t + 1. That is, a prediction is made, at each timepoint, of the location of the target at the next timepoint. For example, if a target is located at (x, y, z) at time t, then it may be predicted that the target will be located at (x', y', z') at time t+1. Where there are two or more targets sensed at timepoint t, a prediction may be made for each of the targets. For example, if there are three sensed targets at time t, with coordinates (xi, yi, zi), (X2, y2, Z2) and (X3, ya, Z3), then the locations of the sensed targets at time t+1 may be predicted to be (x'i , y'1, z'i), (x'2, y'2, z'2) and (x'3, y'3, z'3) respectively. There are numerous ways in which the location of the target at t+1 may be predicted. For example, an Unscented Kalman Filter (UKF) may be used to make the prediction.

[0099] In step 608, the predicted location is compared with the location sensed at time t+1. That is, the predicted t+1 location is compared with the measured t+1 location. A distance between the predicted t+1 location and the sensed t+1 location is calculated. The distance may be a Euclidean distance.

[0100] Optionally, steps 604 and 606 may be performed between the steps of predicting the location at t+1 and calculating a distance between the predicted location and the actual location, especially where there are two or more sensed targets in the indoor environment. When there are two or more sensed targets, it is desirable to correctly match each target to a sensor reading so that the motion of the target can be more accurately monitored. For clarity, steps 602 and 604 will be discussed with reference to a first sensed target and a second sensed target. However, the skilled person would readily understand that the same method may apply to any number of sensed targets.

[0101] At step 604, a distance between each of the predicted locations and each of the sensed locations is determined.

[0102] For example, at time t, locations of a first sensed target p1 and a second sensed target p2 may be sensed as (xi, yi, zi) and (x2, y2, z2). At time t, the predicted locations for the first sensed target and the second sensed target at time t+1 may be (x'i , y'1, z'1) and (x'2, y'2, z'2). At time t+1 , the sensing means may detect a presence of a target at (X3, ys, Z3) and a presence of another target at (x4, y4, z4). At this point, it may not be known whether the target sensed at (x3,ys, Z3) relates to target p1 or target p2 (or even to a new target, p3). Similarly, it may not be known whether the target sensed at (X4, y4, Z4) relates to target p1 , target p2, or a new target.

[0103] The distances between each of the predicted locations and each of the measured locations is calculated. The distances may be Euclidean distances. For example, a distance between the following pairs is calculated:• (x'i, y'i, z'i) and (x3, y3, z3)• (x'i, y'i, z'i) and (x4, y4, z4)• (x'2, y'2, z'2) and (X3, ys, Z3); and• (x'2, y'2, z'2) and (x4, y4, z4).

[0104] In step 606, the sensed locations at time t+1 are assigned to the predicted locations based on the distances calculated in step 604. The predicted locations and sensed locations may be assigned to one another by minimising a cost function. A penalty may be applied to costs above a predetermined distance where it would be unusual for a person to move at the speed necessary to cover that ground in the time between t and t+1. For example, a penalty may be applied to costs of 300cm or more. The cost function may relate to the distances between the predicted locations and the sensed locations. The cost function may relate to one or more attributes instead of, or in addition to, the distances. These attributes may be a reflected power of the sensed target or a height of the sensed target. The cost function may relate to any suitable metric or attribute as understood by the skilled person.

[0105] Where steps 604 and 606 are carried out, the relevant distances calculated in step 604 may be used in step 608. That is, it may not be necessary to calculate again the the distance between the predicted location and the actual location for each target at step 608, if the distances have already been determined in step 604.

[0106] At step 610, the distance calculated in step 608 is compared with a threshold value in order to determine whether the target sensed at time t+1 should be assigned an existing ID or a new ID. If the calculated distance is greater than the threshold value, the sensor reading obtained at t+1 is assigned a new identifier (step 612). If the calculated distance is less than the threshold value, the sensor reading obtained at t+1 is assigned an existing identifier (step 614).

[0107] For example, if a distance between a predicted t+1 location and a measured t+1 location for a sensed target is d1 , then distance d1 is compared with the threshold value. The threshold value may be chosen such that there is a high likelihood that the sensor readings obtained at time t and at time t+1 relate to a same target. The high likelihood may be defined as between a 60%-99% chance that the sensor readings relate to the same target. For example, the high likelihood may be at least 60%, at least 75%, at least 95%, or any other suitable value.

[0108] For example, consider a case where there is a single sensor reading obtained at time t and another single sensor reading obtained at time t+1 The sensor reading obtained at time t is assigned to a target p1. If a predicted location of the target p1 at time t+1 is within 5cm of a location sensed by the sensor at time t+1 , then it may be determined that the target sensed at time t+1 is highly likely to be the same target p1. In this case, the method moves to step 614 and the identifier of target p1 is assigned to the sensor reading at time t+1 .

[0109] However, if the predicted location of the target p1 at time t+1 is 2m away from the location sensed at t+1 , then it may be determined that the target sensed at time t+1 is not target p1 , because it is unlikely that target p1 would have been able to move that distance within that length of time. In this case, the method moves to step 612, and a new identifier is assigned to the sensor reading at t+1 .

[0110] FIG. 7 shows an example of a method 700 for determining whether a sensed target has exited the indoor environment. The sensed target may be a person (or true target) who has been sensed by the sensing means in the room. That is, FIG. 7 shows at least some of the steps which may be performed when carrying out step 504 of FIG. 5.

[0111] In step 502, an exit location of the sensed target is obtained. The exit location is a location within the boundaries of the indoor environment, and may be referred to as an end location or final location. The exit location is a location in the room where the sensed target is last sensed. That is, the exit location may be x, y, z coordinates of a final sensor reading of the sensed target in the room. The final sensor reading may comprise location information (such as x, y, z coordinates) of the target at a time t, where the target is not sensed in the room for at least a predetermined period of time after time t. For example, the final sensor reading of a first target may occur at a time t, where no further sensor data with coordinates indicating that the first target is in the room is obtained for a period of at least 2s, 5s or 10s after time t. Even if further sensor data relating to the first target is obtained after time t, if the further sensor data indicates that the first target is outside the boundaries of the indoor environment, this data may be disregarded.

[0112] In step 704, the determined exit location is compared with a predetermined location. The predetermined location is a region or area within the indoor environment which is in the vicinity of a valid entryway into and / or out of the room. For example, the predetermined location may be an area within a predetermined distance of a door into the room. The predetermined location and predetermined distance may be the same predetermined location and predetermined distance as that discussed in relation to FIG. 4 when determining whether the sensed target has originated at a valid entry point.

[0113] If the determined exit location is within the predetermined location then it may be determined that the sensed target has exited the room. This is because the last sensed location of the sensed person in the room is within a certain distance of an entry way of the room, so it is assumed that the sensed person thereafter left the room. The status of the sensed target may be set to “exited target”.

[0114] If the determined exit location is outside the predetermined location then it may be determined that the sensed target has not exited the room. When sensed target has not been assigned a new sensor reading for a certain period of time, and the last assigned sensor reading was not within the predetermined location, then the status of the target may be set to “vanished target”.

[0115] FIG. 8A is a photograph of an indoor environment 800 according to an embodiment of the invention. An indoor environment is a room in a building. The indoor environment 800 is a bedroom, such as a bedroom for a patient staying in a mental health facility, but the indoor environment is not limited thereto. For example, the indoor environment may also be a bathroom, toilet room, lounge, kitchen, or other room in a building, especially a building intended for the care of vulnerable people.

[0116] FIG. 8B is an illustration of a first experimental set-up 802a for demonstrating the effectiveness of the present invention at distinguishing between a human target and a false target. FIG. 8C is an illustration of a second experimental set-up 802b for demonstrating the effectiveness of the invention at distinguishing between human and false targets.

[0117] Both experimental set-ups were performed in an indoor environment such as the room of indoor environment 800. The room had floor dimensions of 2.7m by 3.8m and was intended to simulate a bedroom of a patient in a mental health care facility. The room included a door 804 and a sensor 202. A single human participant 808 walked around the room along a path shown by the dashed arrows.

[0118] For the first experimental set-up 802a, a mirror 806 was located at a back of the room, on an opposite side of the room to the door 804. The mirror 806 generated reflected targets of the participant 808.

[0119] For the second experimental set-up 802b, a blocker 810 was placed on a left-hand side of the sensor 202 to create a blind spot in the range of the sensor 202. The blocker 810 comprised a sheet of metal foil.

[0120] The sensor 202 was a mmWave 60GHz multiple input multiple output (MIMO) sensor. The sensor was positioned 2.1 m above the floor, on a wooden arm mounted to one of the walls. This type of sensor was selected for its benefits of a wide field of view, built-in algorithms fortarget identification, and compatibility with common wireless communication protocols (HTTP and MQTT). The sensor output comprised a list of X, Y, Z coordinates at each timepoint of a series of timepoints. Sensing frequency was set at 200ms and data packets containing the sensor data were transmitted via MQTT to a Rock4 SE single-board computer, where the transmitted data was saved to a log file.

[0121] For each of the first experimental set-up 802a and the second experimental set-up 802b, 10 trials between 14-31 seconds in length were conducted. During each trial, the participant 808 followed a predetermined trajectory around the room. All trials involved only one person. A OnePlus 12 mobile phone camera was used to collect a video feed and the true target location was manually annotated as a source of ground truth rounded to the nearest 0.25m. Gridlines were marked on the floor every 0.5m to simplify target annotation.

[0122] In order to compare the performance of our pipeline output against the raw sensor data, we determined the (pre) number of True Positives, False Positives and False Negatives on each trial. We used this to generate a sensitivity, specificity and F1 score for each trial. We then ran target tracking and motion pattern analysis (MPA) on the data and determined the (post) number of True Positives, False Positives and False Negatives of the output.

[0123] FIG. 9 shows the results obtained from the first experimental set-up 802a. The aim of this experiment was to see whether the reflected targets generated by the mirror could be removed by target tracking and MPA.

[0124] Before target tracking and MPA was applied, the results showed an average prediction across trials of 68%, average recall across trials of 90% and average F1 score across trials of 34%. After applying target tracking and MPA, average prediction accuracy across trials increased to 100% and average F1 score across trials increased to 87%. Despite showing an overall decrease (85%) after tracking and MPA, recall improved in 7 out of the 10 trials.

[0125] However, there was a large decrease in trial 9 (see Fig 3) which negatively impacted the overall mean. This was due to a previously detected reflected target disappearing at the same time that a new true target appeared. This meant that the new true target was assigned to the reflected target’s track and was therefore excluded.

[0126] FIG. 10 shows the results obtained from the second experimental set-up 802b. The aim of this experiment was to see whether target tracking and MPA successfully tracked the target identity, even when the target was not detected by the sensor. Before target tracking and MPA was applied, the results showed an average prediction across trials of 64%, average recall across trials of 58% and average F1 score across trials of 32%. After applying target tracking and MPA, average prediction accuracy across trials increased to 82%, average recall across trials increased to 74% and average F1 score across trials increased to 74%.

[0127] Two trials showed worse performance than others (trial 18 and trial 20). In both cases this occurred due to a true target with an extremely low velocity having a high correlation with a stationary false target, which led to the true target being incorrectly classified as a reflection.

[0128] A third experiment (not shown) looked at whether the Kalman prediction yielded better target detection accuracy than the raw sensor data. To do this, we calculated D1 as the Euclidian distance at each timepoint between the ground truth and the nearest sensor target, and D2 as the Euclidian distance at each timepoint between the ground truth and the Kalman prediction.

[0129] Descriptive statistics indicated that there was a difference between D1 (mean=44.20, SD=31 .08) and D2 (mean=39.86, SD=26.60). A Shapiro-Wilks test indicated that the differences were not normally distributed (p<0.001 ). In assessing the effectiveness of the Kalman filter on reducing the distance errors, a Wilcoxon Signed-Rank Test was conducted. The analysis revealed a statistically significant decrease in distance error after applying the Kalman filter (W=19920, df=317, p<0.001 ).

[0130] The term 'processor' is to be interpreted broadly to include a CPU, processing unit, ASIC, logic unit, or programmable gate array etc, and may refer to a single processor or a combination of several processors. Certain aspects of the disclosure may be implemented using machine-readable instructions which may, for example, be executed by a general purpose computer, a special purpose computer, an embedded processor or processors of other programmable data processing devices to realize the functions described in the description and diagrams. In particular, a processor or processing apparatus may execute the machine-readable instructions. Thus, functional modules of the apparatus and devices may be implemented by a processor executing machine readable instructions stored in a memory, or a processor operating in accordance with instructions embedded in logic circuitry. The functional modules may be implemented in a single processor or divided amongst several processors.

[0131] It will be appreciated that embodiments of the present invention can be realised in the form of hardware, software or a combination of hardware and software. Any such software may be stored in the form of volatile or non-volatile storage such as, for example, a storage device like a ROM, whether erasable or rewritable or not, or in the form of memory such as, for example, RAM, memory chips, device or integrated circuits or on an optically or magnetically readable medium such as, for example, a CD, DVD, magnetic disk or magnetic tape. It will be appreciated that the storage devices and storage media are embodiments of machine-readable storage that are suitable for storing a program or programs that, when executed, implement embodiments of the present invention. Accordingly, embodiments provide a program comprising code for implementing a system or method as claimed in any preceding claim and a machine readable storage storing such a program. Still further, embodiments of the present invention may beconveyed electronically via any medium such as a communication signal carried over a wired or wireless connection and embodiments suitably encompass the same.

[0132] Throughout the description and claims of this specification, the words “comprise” and “contain” and variations of them mean “including but not limited to”, and they are not intended to (and do not) exclude other moieties, additives, components, integers or steps. Throughout the description and claims of this specification, the singular encompasses the plural unless the context otherwise requires. In particular, where the indefinite article is used, the specification is to be understood as contemplating plurality as well as singularity, unless the context requires otherwise.

[0133] Features, integers, characteristics, compounds, chemical moieties or groups described in conjunction with a particular aspect, embodiment or example of the invention are to be understood to be applicable to any other aspect, embodiment or example described herein unless incompatible therewith. All of the features disclosed in this specification (including any accompanying claims, abstract and drawings), and / or all of the steps of any method or process so disclosed, may be combined in any combination, except combinations where at least some of such features and / or steps are mutually exclusive. The invention is not restricted to the details of any foregoing embodiments. The invention extends to any novel one, or any novel combination, of the features disclosed in this specification (including any accompanying claims, abstract and drawings), or to any novel one, or any novel combination, of the steps of any method or process so disclosed.

[0134] Unless stated otherwise, the numbers and values provided in the examples discussed throughout the description are included to assist with the understanding of the specification. The numbers used, and any relationships between the numbers, are merely examples and are not intended to be limiting.

[0135] The reader's attention is directed to all papers and documents which are filed concurrently with or previous to this specification in connection with this application and which are open to public inspection with this specification, and the contents of all such papers and documents are incorporated herein by reference.

Claims

CLAIMS1. A computer-implemented method for tracking a human target in an indoor environment, the method comprising: receiving sensor data indicative of a location, at a plurality of timepoints, of a first sensed target, the sensor data obtained by a sensing means and comprising coordinate information of the first sensed target at each timepoint of the plurality of timepoints; analysing a motion of the first sensed target, based on the received sensor data; and classifying the first sensed target as a human target or a false target based on a result of the analysis of the motion.

2. The computer-implemented method of claim 1 , wherein analysing a motion of the first sensed target comprises: extracting one or more motion features indicative of a characteristic of a movement of the first sensed target from the received sensor data; and comparing the extracted one or more features to a predetermined model of expected motion patterns of one or more sensed target classifications in the indoor environment.

3. The computer-implemented method of claim 2, wherein the extracted one or more motion features comprises an initial location of the first sensed target, the initial location being within the indoor environment; and wherein analysing the motion of the first sensed target further comprises: comparing the initial location of the first sensed target to a predetermined location of one or more entryways into the indoor environment; and classifying the first sensed target as the human target or the false target based on a result of the comparison.

4. The computer-implemented method of claim 2 or 3, wherein the extracted one or more motion features comprises a range of movement of the first sensed target, and wherein analysing the motion of the first sensed target further comprises: comparing the range of movement of the first sensed target to a predetermined threshold value; and classifying the first sensed target as the human target or the false target based on a result of the comparison.

5. The computer-implemented method of any one of claims 2 to 4, further comprising receiving sensor data indicative of a location of a second sensed target at the plurality of timepoints;wherein the extracted one or more motion features comprises a velocity coefficient of the first sensed target, wherein the velocity coefficient relates to a ratio of a velocity of the first sensed target compared to a velocity of the second sensed target; and classifying the first sensed target as the human target or the false target comprises comparing the velocity coefficient of the first sensed target with a second predetermined threshold value and classifying the first sensed target based on a result of the velocity coefficient comparison.

6. The computer-implemented method of any one of claims 2 to 5, further comprising receiving sensor data indicative of a location of a third sensed target at the plurality of timepoints; wherein the extracted one or more motion features comprises a distance of the first sensed target from the sensing means; and wherein classifying the first sensed target as the human target or the false target comprises comparing the distance of the first sensed target with a distance of the third sensed target from the sensing means and classifying the first sensed target based on a result of the distance comparison.

7. The computer-implemented method of any one of claims 1 to 6, further comprising assigning a continuous target identifier, ID, to each sensor reading corresponding to the first sensed target.

8. The computer-implemented method of claim 7, wherein assigning the continuous target ID comprises: predicting, for time t, a location of the first sensed target at time t+1 based on the sensed location of the first target at time t; calculating a distance between a sensed reading at t+1 and the predicted location; assigning an existing ID to the first sensed target at t+1 if the calculated distance is less than a predetermined threshold value; and assigning a new ID to the first sensed target at t+1 if the calculated distance is greater than the predetermined threshold value.

9. The computer-implemented method of claim 8 wherein, in a case that there are a plurality of sensed readings at t, and a plurality of sensed readings at t+1 , assigning the continuous target ID further comprises: determining a distance between each predicted location and each sensed reading at t+1 ; andassigning each predicted location to one of the plurality of sensed readings at t+1 by minimising a cost function based on the determined distances between the predicted locations and the sensed readings and a threshold penalty.

10. The computer-implemented method of claim 8 or 9, wherein Unscented Kalman Filtering, UKF, is used to predict the location of the target at t+1 .

11. The computer-implemented method of any one of claims 7 to 10, wherein a Kuhn-Munkres algorithm is used to assign the continuous target ID to the first sensed target.

12. The computer-implemented method of any one of claims 1 to 11 , further comprising determining whether the first sensed target has exited the indoor environment.

13. The computer-implemented method of claim 12, wherein determining whether the first sensed target has exited the indoor environment comprises: determining an exit location of the first sensed target; wherein the exit location is within the indoor environment and comprises coordinate information of the first sensed target obtained at a last time point of the plurality of time points where the first sensed target is still within the indoor environment; comparing the exit location of the first sensed target with a predetermined location of one or more exits out of the indoor environment; and determining that the first sensed target has exited the indoor environment if the exit location of the sensed target is within the predetermined location of one of the one or more exits out of the indoor environment.

14. The computer-implemented method of claim 12 or 13, further comprising deleting the sensing data corresponding to the first sensed target from a memory if no further sensor readings of the first sensed target are obtained in the indoor environment for at least a predetermined time period after the first sensed target has exited the indoor environment.

15. The computer-implemented method of any one of claims 1 to 13, further comprising deleting the sensing data corresponding to the first sensed target from a memory if the first sensed target is classified as the false target.

16. An apparatus for tracking a human target in an indoor environment, the apparatus comprising: a communication module configured to receive sensor data indicative of a location of a first sensed target at a plurality of timepoints, the sensor data obtained by a sensing means andcomprising coordinate information of the first sensed target at each timepoint of the plurality of timepoints; a memory configured to store the received sensor data; and a processor; wherein the processor is configured to: analyse a motion of the first sensed target, based on the received sensor data; and classify the first sensed target as a human target or a false target based on a result of the analysis of the motion.

17. The apparatus of claim 16, wherein the processor is configured to analyse the motion of the first sensed target by: extracting one or more motion features indicative of a characteristic of a movement of the first sensed target from the received sensor data; and comparing the extracted one or more features to a predetermined model of expected motion patterns of one or more sensed target classifications in the indoor environment.

18. The apparatus of claim 17, wherein the extracted one or more motion features comprises an initial location of the first sensed target, the initial location being within the indoor environment; and wherein the processor is further configured to analyse the motion of the first sensed target by: comparing the initial location of the first sensed target to a predetermined location of one or more entryways into the indoor environment; and classifying the first sensed target as the human target or the false target based on a result of the comparison.

19. The apparatus of claim 17 or 18, wherein the extracted one or more motion features comprises a range of movement of the first sensed target, and wherein the processor is further configured to analyse the motion of the first sensed target by: comparing the range of movement of the first sensed target to a predetermined threshold value; and classifying the first sensed target as the human target or the false target based on a result of the comparison.

20. The apparatus of any one of claims 17 to 19, wherein the processor is further configured to receive sensor data indicative of a location of a second sensed target at the plurality of timepoints;wherein the extracted one or more motion features comprises a velocity coefficient of the first sensed target, wherein the velocity coefficient relates to a ratio of a velocity of the first sensed target compared to a velocity of the second sensed target; and wherein classifying the first sensed target as the human target or the false target comprises comparing the velocity coefficient of the first sensed target with a second predetermined threshold value and classifying the first sensed target based on a result of the velocity coefficient comparison.

21. The apparatus of any one of claims 17 to 20, wherein the processor is further configured to receive sensor data indicative of a location of a third sensed target at the plurality of timepoints; wherein the extracted one or more motion features comprises a distance of the first sensed target from the sensing means; and wherein classifying the first sensed target as the human target or the false target comprises comparing the distance of the first sensed target with a distance of the third sensed target from the sensing means and classifying the first sensed target based on a result of the distance comparison.

22. The apparatus of any one of claims 16 to 21 , wherein the processor is further configured to assign a continuous target identifier, ID, to each sensor reading corresponding to the first sensed target.

23. The apparatus of claim 22, wherein the processor is configured to assign the continuous target ID by: predicting, for time t, a location of the first sensed target at time t+1 based on the sensed location of the first target at time t; calculating a distance between a sensed reading at t+1 and the predicted location; assigning an existing ID to the first sensed target at t+1 if the calculated distance is less than a predetermined threshold value; assigning a new ID to the first sensed target at t+1 if the calculated distance is greater than the predetermined threshold value.

24. The apparatus of claim 23, wherein, in a case that there are a plurality of sensed readings at t, and a plurality of sensed readings at t+1 , the processor is further configured to assign the continuous target ID by: determining a distance between each predicted location and each sensed reading at t+1 ;assigning each predicted location to one of the plurality of sensed readings at t+1 by minimising a cost function based on the determined distances between the predicted locations and the sensed readings and a threshold penalty.

25. The apparatus of any one of claims 16 to 23, wherein the processor is further configured to determine whether the first sensed target has exited the indoor environment.

26. The apparatus of claim 25, wherein the processor is configured to determine whether the first sensed target has exited the indoor environment by: determining an exit location of the first sensed target; wherein the exit location is within the indoor environment and comprises coordinate information of the first sensed target obtained at a last time point of the plurality of time points where the first sensed target is still within the indoor environment; comparing the exit location of the first sensed target with a predetermined location of one or more exits out of the indoor environment; and determining that the first sensed target has exited the indoor environment if the exit location of the sensed target is within the predetermined location of one of the one or more exits out of the indoor environment.

27. The apparatus of claim 25 or 26, wherein the processor is further configured to delete the sensing data corresponding to the first sensed target from the memory if no further sensor readings of the first sensed target are obtained in the indoor environment for at least a predetermined time period after the first sensed target has exited the indoor environment.

28. The apparatus of any one of claims 16 to 27, wherein the processor is further configured to delete the sensing data corresponding to the first sensed target from the memory if the first sensed target is classified as the false target.

29. A system for tracking a human target in an indoor environment, the system comprising: one or more radio frequency sensors configured to obtain sensor data indicative of a sensed human target in the indoor environment; and the apparatus according to any one of claims 16 to 28.

30. Computer software which, when executed, is arranged to perform a method according to any of claim 1 to claim 15.

Images