Computer-implemented method for pressure sore monitoring

Abstract

The invention relates to a computer-implemented method for pressure sore monitoring. The aim of continuously monitoring both inpatient areas and home areas and of efficiently and effectively evaluating the acquired data in order to provide pressure sore prophylaxis is achieved in that received reflection signals of transmitted radar signals are processed and evaluated such that a change in the pose of a person over time is monitored.

Description

[0001] Computer-implemented method for pressure ulcer monitoring

[0002] The invention relates to a computer-implemented method for pressure ulcer monitoring (e.g. using a sensor arrangement).

[0003] Ongoing assessment of health and well-being, particularly at home, can be an effective way to detect early signs of the onset of conditions so that early medical intervention can be offered, reducing and in many cases avoiding the need for hospital admission.

[0004] In view of the growing nursing shortage due to a lack of staff in the healthcare system, the resulting shortage of beds in inpatient care, but also the ever-increasing life expectancy of people and the resulting aging of the population, the need for ongoing monitoring in both inpatient and home care appears increasingly necessary.

[0005] Continuous monitoring of various health parameters that indicate the state of health can help to detect diseases early or to make a reasonable risk assessment for a person who, for example, develops chronic heart failure or the like.

[0006] Continuous monitoring of older people in their familiar home environment can also help them remain independent for longer, as relatives or emergency services can be immediately notified in an emergency to assist the person. Especially when older people live alone and, for example, fall in their home, they are often unable to call for help themselves and may not be found until days later.

[0007] Previous systems for monitoring indoor spaces require the use of a large number of detection units, transmitting and receiving units, and additional equipment, which makes costs very high. US 2021 / 141082 A1 discloses a system and method for generating and / or updating an indoor environment map. An indoor environment is detected using one or more sensors, and the indoor environment map is generated and updated from the processed data of the environment. 3D radar technology is used to detect the indoor environment. Radio waves are capable of detecting objects in poor visibility conditions such as darkness, smoke, haze, and fog. The system from US 2021 / 141082 A1 can create a three-dimensional (3D) voxel map of the environment and update it in real time. Each voxel is assigned a timestamp and, optionally, respective event characteristics.

[0008] If people lie in a fixed or barely changed position for an extended period of time (e.g., in their home, a hospital, a care facility, etc.), this can lead to a decubitus ulcer (also known as a bedsore or pressure ulcer). To prevent this, regular repositioning may be necessary.

[0009] The invention described below is based on the object of providing a way to continuously monitor both inpatient and home environments (for pressure ulcer prevention) and to efficiently and effectively evaluate the recorded data in order to determine whether and when repositioning a person is necessary. Clearly, pressure ulcer monitoring is provided according to various aspects, thus enabling pressure ulcer prevention. Since it can be determined whether repositioning is even necessary, repositioning at set intervals can be dispensed with, for example, which can reduce the burden on staff, caregivers, etc.

[0010] In this regard, it is also an object of the present invention to provide a method and a device which can be easily adapted to different applications.

[0011] This object is achieved by a computer-implemented method for pressure ulcer monitoring according to independent claim 1. The method according to the invention runs on a microprocessor and comprises the following steps:

[0012] - Receiving data that is acquired by means of a transmitting and receiving unit, wherein the transmitting and receiving unit is designed to transmit radar signals into the monitoring area and to receive their reflection signals from the monitoring area,

[0013] - Calculating distance data from the received reflection signals,

[0014] - Calculate a reflection strength for each three-dimensional (3D) position in the

[0015] Monitoring area from the received reflection signals,

[0016] - Determine valid reflection values ​​from the reflection strength for each 3D position in the monitoring area and create a point cloud,

[0017] - Grouping individual points of the point cloud to an object and tracking a position of the object in the surveillance area, whereby

[0018] - for the presence detection of an object, an X / Y position in the monitoring area is detected and monitored, and

[0019] - for a fall detection of the object, in addition to an X / Y position in the monitoring area, both a spatial and temporal monitoring of a Z position of the object is detected and monitored, and

[0020] - for a measurement of vital parameters of the object, phase values ​​of the determined reflection signals for this object are evaluated and linked to the values ​​of the presence and fall detection of the object.

[0021] Pressure ulcer monitoring, as used herein, can be understood as monitoring a position of a person's body (e.g., a position of the body, a posture (e.g., whether the person is lying on their back, side, or stomach, or whether the person is sitting in bed, etc.), an orientation, etc.). Monitoring the position of the person's body can be understood as meaning that a change in position can be detected. In some aspects, this can be done by determining the explicit position of the body, and by comparing it with a previous explicit position of the body, it can then be determined whether the position has changed (e.g., to what extent it has changed) or not. In other aspects, a change in the position of the body can be detected without determining the explicit position of the body (e.g., an explicit posture).This has the advantage, for example, that less computational and / or memory effort is required to determine the change in body position. Determining the change in position in this case is also less prone to errors, since when determining the explicit position, similar positions could be classified the same, and it could then be determined that no change in position has occurred (e.g., if the person was previously lying on their back and then a lateral position of approximately 30° is classified as a supine position).

[0022] As an example, the change in the position of the body can be determined (without determining the explicit location) by determining a profile and / or heat map based on the intensity of the static reflections (e.g., due to a change in intensity greater than a threshold) and comparing it with a previous profile or heat map (e.g., using a correlation method). If the difference between these is greater than a corresponding threshold, it can be determined that the location has changed.

[0023] In another example, the explicit position can be determined by inputting the point cloud (and optionally further information about other people in the room) into a machine learning model that is configured to output a position (e.g., supine, lateral, etc.) of the person in response to the input.

[0024] In a first step, the transmitting unit of a sensor array transmits radar signals in the 60-64 GHz range into a surveillance area via a phased-array antenna. Other frequency ranges, such as 120 GHz, are also feasible. The phased-array antenna consists of at least three transmitting antennas and one receiving antenna, but more than one receiving antenna can also be used. The reflected signals, which are reflected back to the sensor array, are received by a receiving unit of the sensor array. An in-phase and quadrature method is used before digitizing the signals to capture both amplitude and phase values, which are then evaluated.To increase the signal-to-noise ratio and thus better / more accurate / more precise detection of the amplitude and phase values, radar signals are repeatedly transmitted and received before the received data is evaluated for a processing cycle.

[0025] Distance data is calculated from the received reflection signals. For this purpose, the surveillance area is divided radially into distance sections starting from the sensor. The extent of each distance section is a few centimeters, 3 cm is advantageous. This depends in particular on the bandwidth of the transmitter unit used. By frequency analysis of the reflection signals, an amplitude and phase value is calculated for each distance section. The amplitude value is a measure of the reflection strength of the signals reflected by the objects in the distance section. The phase value is a measure of the distance of an object within a distance section. However, the phase value does not represent an absolute value for a distance. Due to the short wavelength of the radar signals, in contrast to the longer distance section, several distances can correspond to the same phase value.Therefore, only changes in distance can be measured / detected. If an object moves within a certain distance range, the changes can be detected with an accuracy in the micrometer range, for example, with an accuracy of 30 pm.

[0026] It is advantageous to exclude static objects from the received reflection signals, as these are stationary and irrelevant for pressure ulcer monitoring. Due to the division of the monitoring area into distance segments described here, in conjunction with the evaluation of the phase values ​​of the radar signals, object movements in the millimeter and micrometer range, such as movement of the person's torso during breathing and / or due to the heartbeat, can be detected. This allows living beings to be distinguished from static objects.

[0027] This simplifies and accelerates the process. It is therefore particularly advantageous if the received reflection signals are preconfigured so that objects with different movement intensities can be differentiated from the reflection signals. To do this, the data from the received reflection signals are processed using various calculation rules, the resulting data is saved and then further processed. This allows the results to be combined later. The preconfiguration of the data can, for example, be such that static objects are calculated out of the data using an initial calculation rule. Static objects are objects that exhibit no movement whatsoever. This makes it easier to distinguish living beings from other objects, for example. The resulting data simplifies and accelerates the detection of the position of the monitored person.Using a second calculation rule, object movements are subtracted in order to detect static objects in the room. This enables mapping of the monitored room and is useful, for example, for detecting the presence of an object / person in the bed and / or in the room. Using a further calculation rule, the resulting data after the first calculation rule is recombined over several processing cycles. This subtracts out rapid, large movements, such as walking or falling of living beings, if any, and amplifies fine, slow movements, such as breathing and speaking. This is useful for detecting living beings when they are very still, for example while sleeping or after a fall. This information can be used to clearly determine, for example, whether there is a person in the monitored area who has been lying in a bed for an extended period of time.For this person, for example, the pressure ulcer monitoring described here can be carried out by monitoring the position of the ulcer.

[0028] The data is processed separately for each distance section, which is also referred to as a distance range.

[0029] In a further step, a reflection strength is calculated from the received reflection signals for each position in the three-dimensional space of the surveillance area. Starting with a spherical coordinate system, each distance range is further divided into angular ranges for the azimuth angle and elevation angle. Using a spectrum-based beamforming process, the reflection strength is calculated for each of these spatial areas defined by distance and solid angle. Valid reflection values ​​are determined from the reflection strength for each 3D position in the surveillance area, and a point cloud is created from them. A valid reflection value is defined as the signal strength of the received reflection signals originating from a target object. These can vary greatly, depending, for example, on the distance and the surface material of the object.To avoid this problem, the threshold for a valid reflection value is recalculated for each spatial area, also known as a cell, from its local environment. To do this, each cell is compared with neighboring cells. To do this, the sum of the surrounding cells for a specific area is determined. Directly adjacent cells can be ignored. The sum is calculated using a fixed factor to produce a threshold value. The threshold value is compared with the value of the cell under investigation. If the cell value is greater than the threshold, it is a valid reflection value and the spatial area is added to the point cloud as a point. The factor is empirically set to an optimum so that all target objects are detected with a minimum number of false detections. The cells can be evaluated using a combination of 1D and 2D processing, which is sufficiently accurate and fast, or directly using 3D processing.For each acquired point, the position, signal-to-noise ratio, and relative velocity to the sensor are calculated and saved. This is schematically shown in Fig. 1. In a next step, individual points in the point cloud that have similar values ​​regarding position, velocity, and signal-to-noise ratio are grouped into an object. Points can either be assigned to an object known from a previous processing cycle or combined into a new object. Points that cannot be assigned are discarded.

[0030] First, it is checked which points can be assigned to an existing object. This involves two steps. First, the calculation of a threshold value for the distance and second, the calculation of a score for membership. When calculating the distance threshold value, the predicted state vector (location, velocity, acceleration) of the object, the previous distribution of points of this object in space and the measurement uncertainty are taken into account. If a point in the point cloud is closer than the calculated threshold value, it may belong to this object. This can be true for several objects. For an unambiguous assignment, the score for membership is calculated in the next step. For this purpose, the Mahalanobis distance, with increased factorization of the velocity, between a point and each possible object is calculated.Since the objects to be distinguished are already spatially close to each other, amplifying the speed during the calculation provides a better classification. The point is assigned to the object with the highest score.

[0031] If points are combined to form a new object, the values ​​of any point are assumed to be the object center and object speed. Each remaining point is checked one after the other to determine whether it complies with the limits for distance and speed difference. If the limits are met, the new point is added to the object, and the object center and object speed are recalculated to start the comparison again with the next point. After all existing points have been tested, the created object is checked for validity. According to various criteria, a valid new object is saved if the limits for the number of points, the combined signal-to-noise ratio, and the minimum speed are exceeded.

[0032] For example, a grouped object can be a person lying in a bed whose position is being monitored (for pressure ulcer prevention).

[0033] In general, the position of the grouped object in the surveillance area is tracked so that, for example, an X / Y position in the surveillance area can be detected and monitored for the presence detection of an object.

[0034] Furthermore, in addition to the X / Y position in the surveillance area, both spatial and temporal monitoring of the Z position of the object can be detected and monitored. The center of gravity of the point cloud of all points of an object can be the used Z position. However, the minimum / maximum Z position (lowest / highest point) can also be included in the evaluation. For example, the Z position fluctuates considerably when walking, as different body parts reflect with different strengths in each processing cycle. If the person / object is lying down (e.g. in a bed or on the floor), no matter where reflections occur, the center of gravity will always be at a low height. To verify and fine-tune the recorded data, the object's vital parameters can also be measured; for this purpose, phase values ​​of the recorded reflection signals of this object are evaluated.As explained herein, vital signs can be used to distinguish persons (as dynamic objects) from static objects even when the person is, for example, lying in a bed and sleeping.

[0035] The phase values ​​represent object movements in the millimeter and micrometer range. In humans, the skin reflects the radar signals. Skin movement, caused by breathing and heartbeat, causes a change in phase. The change in phase is directly proportional to the change in distance caused by the skin movement (Fig. 2).

[0036] In one embodiment of the computer-implemented method according to the invention, conclusions about the presence of the object are drawn using a heat signature at an object position, which is detected using a thermopile array. An object to be verified is located at an object position. If the thermopile array detects thermal radiation that is higher than the object's surroundings, it can be deduced that the object at this position may be a living being, e.g., a human or an animal. However, this assumption must be further specified.

[0037] In a further embodiment of the method according to the invention, a fine motion test based on phase values ​​is used to detect whether the detected object is moving, allowing a reliable statement to be made as to whether it is a living being or not. The phase values ​​are extracted from the received reflection signals for each 3D position in the surveillance area and linked to the grouped object.

[0038] In another embodiment of the method according to the invention, the fine movement test for determining vital parameters comprises the following steps, whereby in particular a respiratory rate of the object can be determined:

[0039] - Selecting an object position and the corresponding reflection signals and

[0040] Phase values ​​in a spatial area,

[0041] - Filtering the reflection signals with an IIR bandpass filter for the

[0042] Respiratory rate range,

[0043] - Performing a frequency analysis using Fourier transformation and

[0044] Calculating an amplitude spectrum for the respiratory rate range,

[0045] - Determination of the highest amplitude values ​​of the amplitude spectrum as well as their corresponding frequency values ​​and background noise,

[0046] - Sorting the determined amplitude values ​​according to their signal-to-noise ratio, whereby a predetermined percentage of the determined amplitude values ​​is selected and used for further processing,

[0047] - Calculating a median value of all frequency values ​​of the selected and used amplitude values, wherein the median value represents a value for the respiratory rate. In a further embodiment of the method according to the invention, the fine movement test for determining vital parameters comprises the following steps, whereby, in particular, a heart rate of the subject can be detected:

[0048] - Selecting an object position and the corresponding reflection signals as well as

[0049] Phase values ​​in a spatial area,

[0050] - Filtering the reflection signals with an IIR bandpass filter for the

[0051] Frequency range of the heart rate,

[0052] - Performing a frequency analysis using Fourier transformation and

[0053] Calculating an amplitude spectrum for the frequency range of the heart rate,

[0054] - Cleaning the calculated amplitude spectrum of interference from harmonic frequencies of the respiratory rate and random object movements by detecting harmonic frequencies of the potential heart rate and storing a weighted fundamental frequency value in a data buffer,

[0055] - Perform a cluster analysis on the data stored in the data buffer

[0056] Frequencies, where a frequency average of the largest cluster is stored,

[0057] - Repeat the above processing step and calculate a

[0058] Median value of the frequency averages stored over several processing cycles, where the median value represents a value for the heart rate.

[0059] Body functions such as respiratory rate or heart rate exhibit a periodicity in different frequency ranges, which is advantageously used according to the invention for determining these vital parameters.

[0060] Using the object position, the correct distance sections with their phase values ​​are selected.

[0061] The phase values ​​are stored in successive processing cycles. Phase values ​​are available for each transmitting and receiving antenna pair at each distance. First, these signals are filtered with an IIR bandpass filter. Two different filters are used: one for the respiration frequency range and one for the heartbeat frequency range to determine the heart rate. The cutoff frequencies of the filters are selected so that typical values ​​for heart and respiration rate are not filtered. To create a frequency analysis of the filtered data, a discrete Fourier transform is then applied to the data. The result of the Fourier transform is used to calculate the amplitude spectrum for the corresponding frequency ranges. For evaluation, the highest amplitude values ​​and their corresponding frequency value, as well as the noise floor, are determined.These values ​​are determined and compared for all antenna pairs in the selected distance sections.

[0062] For respiration, all amplitude values ​​are sorted according to their signal-to-noise ratio. Starting with the highest signal-to-noise ratio, only a certain percentage is used for further processing. The median is calculated from all frequency values ​​of these selected amplitude values. These values ​​are stored over several processing steps, and the median value is calculated in each cycle. This value is the result for the respiratory rate.

[0063] When evaluating the heart rate spectrum, interference caused by harmonic frequencies of respiration and random body movements should be taken into account. These also cause high amplitudes in the frequency range of the heart rate. The presence of high amplitude values ​​at harmonic frequencies has proven to be a reliable distinguishing feature. A harmonic frequency is considered present if a limit for the signal-to-noise ratio is exceeded in a frequency range by an integer multiple of the fundamental frequency of an amplitude value. Interference also has high amplitude values ​​in the heart rate range (0.8 Hz - 2.0 Hz), but not high values ​​in the harmonic range (1.6 Hz - 6.0 Hz). Only if no harmonic frequencies are found, then the frequency value transferred to the data buffer may not be a multiple of the respiratory rate.When determining the heart rate, the calculated amplitude spectrum must therefore be cleaned of these disturbing harmonic frequencies of respiration.

[0064] In another further embodiment of the method according to the invention, only frequency values ​​that satisfy the following rules are transferred to the data buffer when cleaning the calculated amplitude spectrum:

[0065] - If a 1st and 2nd harmonic of the selected frequency value is present, the frequency value is transferred four times into the data buffer,

[0066] - If the 1st or 2nd harmonic of the selected frequency value is present, the frequency value is transferred three times to the data buffer,

[0067] - If the amplitude value is the highest in the amplitude spectrum and the corresponding frequency value is not a harmonic of the respiratory rate, the frequency value is adopted twice,

[0068] - If the difference between a frequency value and one of the three highest amplitude values ​​in the amplitude spectrum and a result for the heart rate from a previous processing cycle is smaller than a limit value, this frequency value is adopted, whereby

[0069] - the rules are checked in the above order until a rule applies.

[0070] All frequency values ​​found for all antenna pairs from all distance sections are checked and transferred to the data buffer. Certain rules apply to whether and how the value is transferred:

[0071] - if the 1st and 2nd harmonic of the selected frequency value is present, the frequency value is adopted four times,

[0072] - if the 1st or 2nd harmonic of the selected frequency value is present, the frequency value is adopted three times,

[0073] - if the amplitude value is the highest in the amplitude spectrum and the frequency value is not a harmonic of the respiratory rate, the value is adopted twice, - if the difference between the frequency value of one of the three highest amplitude values ​​in the amplitude spectrum and the result for the heart rate from a previous processing cycle is smaller than a limit value, the value is simply adopted.

[0074] The rules are checked in the order listed and as soon as one of the rules applies to a frequency value for an antenna pair, no further rules and frequency values ​​are checked for this antenna pair.

[0075] Cluster analysis is applied to the data buffer containing all collected frequency values. Clustering is beneficial for excluding outliers from the averaging and suppressing noise in individual distance segments. The mean of the largest cluster is used. These values ​​are stored over several processing steps, and then the median is calculated. This value is the result for the heart rate.

[0076] In one embodiment of the method according to the invention, a confidence level is defined which is used to check whether calculated values ​​of the vital parameters are plausible or whether they are falsely positive detected vital parameters.

[0077] It has proven advantageous to define a so-called confidence level, which is used to check whether calculated values ​​are plausible or whether they are false positives, which are then discarded. The heart rate result is stored in each journal. The median of these values ​​is then calculated over a specific period of time. The number of values ​​within this period with a difference from the median that is smaller than a defined threshold is then counted. The ratio of this number to the total number of values ​​defines the confidence level.

[0078] If the algorithm has recently determined many different values, the confidence is low; otherwise, it is high. This has the advantage that values ​​from periods in which the sensor is likely to be incorrectly positioned are not included in the averaging of the minute, hour, or daily values.

[0079] In a further embodiment of the method according to the invention, the surveillance area is configured by classifying objects in the surveillance area using an object classification system. The object classification data is used for a self-learning algorithm for creating and continuously improving a spatial configuration of the surveillance area. The algorithm can, for example, be trained to detect the self-movement of living beings.

[0080] For room configuration, data is collected and made available to the algorithm for learning. The more data the algorithm has access to over time, the more accurate and reliable the room configuration can be. For example, the position of a bed is determined by the sensor's self-learning process, based on a person repeatedly lying in the same position. "Person" and "lying" then come from the classification. Room size and the position of doors can be learned in the same way.

[0081] The detected objects and their respective point clouds serve as input data for object classification. Specific features can be derived from the number of spatial points assigned to an object in each processing cycle. The spatial distribution of the points yields the object's length, width, height, and center of gravity. The minimum, maximum, and average speeds are determined from the speed values ​​of the individual points. The features are determined in each processing cycle and, after a certain number of cycles, combined into a feature vector. Depending on the application, features are combined, filtered, or only certain ones are used. A classifier is trained and evaluated using recorded and manually classified data. For this purpose, a cluster analysis of all feature vectors in the feature space is performed.The trained classifier makes decisions based on the distances to the identified cluster centroids. There is one classification that recognizes human objects. Another classification determines posture and distinguishes between lying, sitting, standing, and walking.

[0082] In one embodiment of the computer-implemented method, binary phase modulation is applied to the phased antenna array. Binary phase modulation allows all antennas of the phased antenna array to transmit and receive simultaneously, with the superposition subsequently being removed to process the antenna signals individually. Simultaneous transmission improves the signal-to-noise ratio and thus the accuracy of amplitude and phase.

[0083] In another embodiment of the computer-implemented method, a zoom-in function is used to monitor objects in the surveillance area. The spatial areas containing the detected and evaluated target object are divided into smaller spatial areas in order to determine the object's position more precisely.

[0084] In a further embodiment of the computer-implemented method, a circle approximation is used to detect static objects in the surveillance area. Circle approximation in the complex plane has the advantage that it can more effectively eliminate static objects than the mean value method.

[0085] In another further embodiment of the computer-implemented method, an extended differentiation and cross-multiplication algorithm is used to calculate the phase values. The extended differentiation and cross-multiplication algorithm (Extended DACM) can be used instead of an acrtan function to calculate the phase, allowing for faster calculation of the phase values.

[0086] In a further embodiment of the computer-implemented method, a motion index is assigned to the objects, and the vital parameters are calculated only at low movement intensity. Calculating the vital parameters only at low movement intensity has the advantage of allowing more reliable values ​​to be recorded for the evaluation of the detected object. However, vital parameters are not evaluated for objects in motion.

[0087] The advantage of the method according to the invention is that it can reliably determine whether the detected object is a living being, e.g., a human or an animal, and whether repositioning is necessary for pressure ulcer prevention. Based on the reflection signals from the points associated with a person, the person's position can be determined, so that a temporal change in this position can be used for pressure ulcer monitoring. Based on the determined vital parameters, which are extracted from the received phase values ​​of the reflection signals, the health status of the living being can be derived.

[0088] The object of the invention is also achieved on the arrangement side by a sensor arrangement according to independent claim 11. The sensor arrangement according to the invention for pressure ulcer monitoring can have a transmitting and receiving unit and one or more than one processor configured to carry out the method according to claim 1.

[0089] In one embodiment of the sensor arrangement according to the invention, the sensor arrangement comprises a system-on-chip (SoC), which has, for example, an analog front end, a digital signal processor, a microcontroller and a memory, which are connected to one another in a communicative manner via signal lines, as well as a phase-controlled antenna array, which is designed to emit radar signals and to receive their reflection signals, and which can be controlled by the microcontroller, wherein the received reflection signals can be processed in the digital signal processor and / or the microcontroller.

[0090] In another embodiment of the sensor arrangement according to the invention, it has a thermopile array for detecting a temperature of the object in the monitoring area.

[0091] The use of the thermopile array has the advantage that the acquired thermal image can be used as a plausibility check for the reflected signals from the radar data, and vice versa. Temperature measurements can be made more precise using the distance information from the radar measurement by introducing correction factors. This combination can, for example, distinguish living objects from physical objects, and can also be used to classify living objects, e.g., humans or animals.

[0092] For this purpose, a geometric camera model is calculated by the sensor. With the help of this model, a direct conversion of 3D spatial points from the radar into the 2D pixel plane of the thermopile array can be carried out. The precise assignment of specific pixels and thus also temperatures to an object detected by the radar increases the accuracy of presence and fall detection. Moving objects (e.g., a tipping chair, fan) are excluded from detection due to their insufficient temperature. For this purpose, the sensor calculates a projection matrix that results from the internal and external orientation of the camera (thermopile array). The external orientation is the position of the sensor in space. The translational parameters are known via the configuration, i.e., the positioning of the sensor arrangement in space. The rotational parameters are determined by evaluating an acceleration sensor.For internal orientation, the physical pinhole camera model is assumed, which requires focal length, pixel size, and principal image point. Furthermore, the distortion of the fisheye lens is modeled. The distortion is radially symmetric, and the lens curvature is known to the sensor. The projection matrix is ​​recalculated for each position change determined by the accelerometer (e.g., each time the sensor is started).

[0093] In a further embodiment of the sensor arrangement according to the invention, the space that can be detected by the sensor arrangement has a size of at least 5m x 5m.

[0094] It is advantageous that the sensor arrangement according to the invention utilizes a combined sensor system for detecting and determining vital parameters, which comprises a transmitting and receiving unit for radar radiation and a thermopile array. Vital parameters include, among others, heart rate, heart rate, respiration or respiratory rate, body temperature, facial temperature, and movements. These can be determined using the sensor arrangement according to the invention and the computer-implemented method.

[0095] The analysis leads to a greater degree of accuracy in distinguishing between healthy and sick individuals, and in the detection of an emergency. It enables the recognition of clinical pictures and the differentiation of acute emergencies, such as falls (using fall detection and motion analysis), heart attacks, sepsis, pulmonary embolism, and others. A suitable communication unit transmits the information, preferably wirelessly, to the necessary or appropriate persons and / or institutions.

[0096] It is advantageous that the system according to the invention operates contactlessly and reliably over distances of up to approximately 4 m, whereby the parameters are continuously recorded, ie a continuous monitoring of the parameters is implemented.

[0097] The invention will be explained in more detail below using some exemplary embodiments.

[0098] The drawings show

[0099] Fig. 1 Schematic representation of a sub-step of the computer-implemented method according to the invention for determining valid reflection values;

[0100] Fig. 2 Time course of the phase signal of the recorded respiratory and heart rate at a sampling rate of 20Hz;

[0101] Fig. 3 Comparison of a respiratory rate determined by the method according to the invention and a reference measuring device applied to a human body; Fig. 4 Comparison of a heart rate determined by the method according to the invention using radar measurement and a reference measuring device applied to a human body;

[0102] Fig. 5 Schematic representation of the computer-implemented method according to the invention in a flow chart;

[0103] Fig. 6 Schematic representation of a scenario I: (a) Person walks through the surveillance area and falls, (b) Time course of object detection, (c) Thermopile image during the person's fall, (d) Visualization of the point cloud before and after the fall;

[0104] Fig. 7 Schematic representation of a scenario II: (a) Person lies in bed, stands up and falls, (b) temporal course of object detection, (c) and (d) recorded phase signal and derived respiratory and heart rate;

[0105] Fig. 8 Integration of the sensor arrangement with the method according to the invention into a centralized emergency call system.

[0106] With the sensor arrangement for pressure ulcer monitoring according to the invention, which has a transmitting and receiving unit and one or more processors configured to carry out the computer-implemented method according to the invention, living beings / objects in a monitoring area can be detected and their position determined at a particular point in time. Based on a temporal change in position, it can then be determined whether (active) repositioning (e.g., by nursing staff, a caring relative, etc.) is necessary as part of pressure ulcer prevention.

[0107] The sensor arrangement with integrated evaluation can be used for carrying out bio-scan and / or bio-monitoring by measuring individual variables and their evaluation for integrated consideration in a learning system, ie • Contactless combined measurement of heart rate, respiratory rate, body temperature and movements,

[0108] • Learning system using an integrating algorithm to increase the discrimination between sick and healthy compared to considering individual parameters,

[0109] • Expandable / adaptable evaluation by changing and / or supplementing the algorithms to suit a wide range of clinical pictures and / or integrating additional measurements.

[0110] Figure 3 shows a comparison of a respiratory rate determined by the method according to the invention with a reference measuring device applied to a human body. Figure 4 shows a comparison of a heart rate determined by the method according to the invention using radar measurement with a reference measuring device applied to a human body. The comparisons with the reference measuring devices show that the data recorded by the sensor array and processed by the method are sufficiently accurate to allow conclusions to be drawn, for example, about a person's state of health.

[0111] Figure 5 shows exemplary aspects of the computer-implemented method in overview, as described in detail in the present application.

[0112] For presence, fall, and vital parameter detection, as well as for pressure ulcer monitoring, radar signals in the 60-64 GHz range are transmitted from a transmitting unit 2. These signals are reflected by obstacles / objects 3 in the monitoring area 1 and received again by a receiving unit. Distance data is calculated from the received reflection signals, with a reflection strength being calculated for each 3D position in the monitoring area from the received reflection data. Using a plausibility check, the valid reflection values ​​are determined from the set of calculated reflection values, and these are combined into a point cloud. Individual points of the point cloud are then grouped into an object and classified using an object classification.The object is detected and monitored based on its X / Y position in the monitoring area. Additional spatial and temporal monitoring of the object's Z position can be used to detect whether a fall has occurred. By measuring the object's vital parameters, which are determined and evaluated from the phase values ​​of the recorded reflection signals for this object, it can be determined whether the object is a living being (e.g. a person). By linking this to the values ​​from presence and fall detection, a plausibility check can be carried out to determine whether the object has fallen or whether it is merely an object, e.g. a vase. If it is indeed a living being, conclusions can also be drawn about the creature's state of health.In the context of pressure ulcer monitoring, fall detection can be used, for example, to determine whether the person's position has changed over time in bed or whether the person has fallen out of bed.

[0113] Figure 6 shows an example of a fall detection and the associated sensor arrangement: A person walks through a room and falls (Fig. 6a). It is understood that falling out of bed can also be considered a fall and detected accordingly (see also Figure 7 and the associated description). The sensor arrangement records the movements of the person / object over time. The method according to the invention runs through the following temporal steps, which must be fulfilled for fall detection (Fig. 6b):

[0114] 1 . The Z position of an object is smaller than a threshold;

[0115] 2. The Z position remains below a limit for a certain time;

[0116] 3. Optionally (if a temperature sensor (e.g. thermopile) is present), calculation and verification of the temperature difference between the object and the environment, if the difference is greater than the limit value;

[0117] 4. Intensity of fine movement above a certain height is less than a

[0118] Limit;

[0119] 5. Fall detected, alarm sent. Fig. 6c shows the thermopile image during a fall. Gray pixels are generated by the body heat of the person lying on the ground. The black pixel cross was inserted into the image by the radar chip and represents the 3D position converted into 2D pixel coordinates. These coordinates are used to calculate the temperature difference between the object and its immediate surroundings.

[0120] Fig. 6d) shows, on the one hand, the realistic scene to be detected by the sensor arrangement and, on the other hand, the visualization of the point cloud of the detected object(s) before and after the fall.

[0121] Figure 7 shows a classic application scenario for the computer-implemented method according to the invention and the associated sensor arrangement: A person is lying in bed, gets up, and falls, or the person falls out of bed. The respiratory and heart rate are continuously recorded in bed (Fig. 7b). If the object position changes, the detection is immediately switched to the new position. Phase extraction immediately provides the algorithm with the new phase signals from all virtual antennas. As a result, values ​​for the vital parameters are continuously transmitted again shortly after the fall is reported. Figs. 7c) and 7d) show the phase signal with a brief interruption and the respiratory and heart rate determined from the phase signal before and after the fall (interruption). Figure 8 shows the integration of the sensor arrangement 2 with fall detection into a centralized emergency call system, a) without and b) with a local emergency call system.Accordingly, an information system (e.g. in a hospital, nursing home, in the home environment, etc.) can also be used to inform the patient about the need to reposition the person as part of pressure ulcer monitoring.

[0122] Consequently, the sensor array can be used to detect emergency and fall situations and to determine vital parameters, for example, for emergency calls, telemedicine, contactless remote diagnosis, and health monitoring. These include, for example: • Bio-scanning using contactless screening of individuals for infectious risks.

[0123] • Screening the health status of individuals accessing critical infrastructure and those with medical conditions that are vulnerable, such as those in intensive care or after organ transplants or in similar situations, by measuring vital signs such as heart rate, respiratory rate, and facial temperature. Integrated interpretation of the measurements increases the test's discriminatory power.

[0124] • Health monitoring for contactless measurement of people’s vital parameters, such as heart and respiratory rate and temperature, in conjunction with movement analysis to detect falls, monitor pressure sores, etc.

[0125] The contactless measurement and evaluation of vital parameters such as respiratory rate, pulse rate, temperature, movement allow various applications, such as

[0126] • Automated emergency call

[0127] • Screening for infectious risks, for example for access control in public institutions, authorities, airports, companies, infection hotspots, to risk groups or for pre-selection for tests,

[0128] • Monitoring of vital parameters, for example for the following purposes: a) health monitoring, b) reporting of limit violations, c) detection of emergencies, such as heart attack, embolism, sleep apnea, etc. d) measurement from a distance in the range of 0.3 m to 5 m, for example by arranging the sensor arrangement on a room ceiling (distance approx. 1 - 2 m),

[0129] • Telemedicine with electronic waiting room, for example including reporting of limit value violations, and remote diagnostics,

[0130] • Motion detection for fall detection and / or smart home applications, veterinary medicine, for example also by means of monitoring and / or diagnosis.

[0131] In particular, the evaluation of vital parameters makes it possible to determine whether an object is a living being (e.g. a person) or not.

[0132] According to various aspects, a three-dimensional spatial resolution of a point cloud can be achieved based on the reflection signals recorded at a given point in time. A reflection strength and a phase value can be assigned to each point in the point cloud. According to various aspects, multiple points in the point cloud can be (unambiguously) assigned to an object (e.g., grouped). Based on the phase values, it can then be determined, for example, whether the object is a living being (e.g., a person). Additionally or alternatively, the detection of the object's heat signature, as described herein, can be used for this purpose.

[0133] Using the reflection signals, not only can an average distance between the object (e.g. the person) and the transmitting and receiving unit (2) be determined, but a respective distance between different points on the object and the transmitting and receiving unit (2) can also be determined. For example, a change in the distance between limbs of the person can be determined. For example, a change in the distance between sections of the person's torso can be determined, whereby, for example, a rotation of the torso can be determined. Using the respective distances between the different points on the object at a specific point in time, a position of the person (e.g. a spatial orientation of the person's body) can be determined. Based on changes in distance, a (temporal) change in the position of the person can then be detected.

[0134] According to various embodiments, it can be determined whether the temporal change in the person's position within a predefined period of time is smaller than a predefined position change threshold. If this is the case, it can be concluded that a repositioning of the person has not yet occurred and is therefore necessary.

[0135] According to various embodiments, it can be determined whether another person is within a predefined distance of the person during the temporal change in the person's position. If this is the case, it can be concluded that a caring person initiated the temporal change in position. If the other person is located between the person and the transmitting and receiving unit (2) in a period of time, no reflection signals can be recorded for at least some of the points belonging to the person in this period. However, if a changed position of the person is determined after the period compared to before the period, it can be determined that the other person repositioned the (caring) person.

[0136] In this way, a clear distinction can be made between the person's own movement and an external movement (e.g. by the other person providing care).

[0137] According to various aspects, it can be determined that the person is having an epileptic seizure based on the frequency and / or intensity and / or regularity of the temporal change in the person's position (for example, if violent movements are detected across the entire body for several minutes, optionally with rhythmic movements). Thus, to detect an epileptic seizure, the duration, speed, and area of ​​the movements occurring during the temporal change in position can be recorded and evaluated. According to various aspects, it can be determined whether or not an epileptic seizure is occurring based on the frequencies of the phase values ​​described herein. Computer-implemented method for pressure ulcer monitoring. List of reference symbols

[0138] 1 Monitoring area 2 Transmitting and receiving unit

[0139] 3 object to be detected

[0140] 4 Processing and evaluation unit

Claims

Computer-implemented method for pressure ulcer monitoring Patent claims 1 . A computer-implemented method for pressure ulcer monitoring in a monitoring area (1), comprising: for each time point of a plurality of successive time points: - receiving data representing reflection signals from radar signals transmitted into the surveillance area (1) at that time, - Calculating distance data from the reflection signals, - Calculating a reflection strength for each 3D position in the monitoring area (1 ) from the reflection signals, - Determining valid reflection values from the reflection strength for each 3D position in the monitoring area (1 ) and creating a point cloud, - Grouping individual points of the point cloud into a respective object (3); Determining, using a temporal change of respective phase values of the points grouped to the respective object (3) in the surveillance area (1) and / or a heat signature of the respective object (3), whether the respective object (3) is a person; and if it is determined that the respective object (3) is a person: - for each point in time of the plurality of successive points in time, determining a respective position of the person using the points in the surveillance area (1) grouped to the object (3), - whereby a temporal change in the person's position is determined for pressure ulcer monitoring.

2. Computer-implemented method according to claim 1, wherein by means of a fine motion test based on the respective phase values obtained for each grouped object (3) from the received reflection signals for each 3D Position in the surveillance area (1 ) is extracted, it is detected whether the object (3) is a person or not.

3. A computer-implemented method according to claim 3, wherein the fine control test comprises the following steps for determining vital parameters, in particular a respiratory rate: - Selection of an object position and the corresponding reflection signals as well as phase values in a spatial area, - Filtering the reflection signals with an IIR bandpass filter for the Respiratory rate range, - Performing a frequency analysis using Fourier transformation and Calculating an amplitude spectrum for the respiratory rate range, - Determination of the highest amplitude values of the amplitude spectrum as well as their corresponding frequency values and background noise, - Sorting the determined amplitude values according to their signal-to-noise Ratio, whereby a predetermined percentage of the determined amplitude values is selected and used for further processing, - Calculating a median value of all frequency values of the selected and used amplitude values, where the median value represents a value for the respiratory rate.

4. A computer-implemented method according to claim 3, wherein the fine movement test comprises the following steps for determining vital parameters, in particular a heart rate: - Selecting an object position and the corresponding reflection signals as well as phase values in a spatial area, - Filtering the reflection signals with an IIR bandpass filter for the Frequency range of the heart rate, - Performing a frequency analysis using Fourier transformation and Calculating an amplitude spectrum for the frequency range of the heart rate, - Cleaning the calculated amplitude spectrum of disturbances by harmonic frequencies of the respiratory rate and random object movements by detecting harmonic frequencies of the potential heart rate and storing a weighted fundamental frequency value in a data buffer, - Performing a cluster analysis on the frequencies stored in the data buffer, storing a frequency average of the largest cluster, - Repeating the aforementioned processing step and calculating a median value of the frequency averages stored over several processing cycles, wherein the median value represents a value for the heart rate.

5. A computer-implemented method according to claim 5, wherein during the cleaning of the calculated amplitude spectrum only frequency values are transferred into the data buffer that satisfy the following rules: - If a 1st and 2nd harmonic of the selected frequency value is present, the frequency value is transferred four times into the data buffer, - If the 1st or 2nd harmonic of the selected frequency value is present, the frequency value is transferred three times into the data buffer, - If the amplitude value is the highest in the amplitude spectrum and the corresponding frequency value is not a harmonic of the respiratory rate, the frequency value is adopted twice, - If the difference between a frequency value and one of the three highest amplitude values in the amplitude spectrum and a result for the heart rate from a previous processing cycle is smaller than a limit value, this frequency value is adopted, whereby - the rules are checked in the above order until a rule applies.

6. Computer-implemented method according to one of the preceding claims, where a confidence level is defined which is used to check whether calculated values of the vital parameters are plausible or whether they are false positive detected vital parameters.

7. Computer-implemented method according to one of the preceding claims, wherein a zoom-in function is used for monitoring objects in the monitoring area.

8. Computer-implemented method according to one of the preceding claims, wherein a movement index is assigned to the objects (3) and the calculation of the vital parameters takes place only at a low movement intensity.

9. Computer-implemented method according to one of the preceding claims, wherein for pressure ulcer monitoring the following is determined: - whether the temporal change in the person's position within a predefined period of time is less than a predefined position change threshold, and - if it is determined that the temporal change in the position of the person within the predefined period is less than the predefined position change threshold, that a repositioning of the person is necessary.

10. Computer-implemented method according to one of the preceding claims, wherein for the purpose of monitoring the person's pressure ulcers, using the points of the point cloud grouped into a respective object (3), it is further determined: - whether another person is within a predefined distance from the person during the temporal change of the person's position, and - if it is determined that during the temporal change of the person's position another person is within a predefined distance from the person, the person has been repositioned; - wherein preferably, if it is determined that during the temporal change of the position of the person the other person is within a predefined distance from the person, the relocation of the person is only determined if the temporal change of the position of the person is greater than or equal to a predefined position change threshold value.

11. A sensor arrangement for pressure ulcer monitoring, comprising: one or more than one processor configured to carry out the method according to claim 1; and the transmitting and receiving unit (2) configured to acquire the data by transmitting radar signals into the monitoring area (1) and receiving their reflection signals from the monitoring area (1).

12. Sensor arrangement according to claim 14, wherein the sensor arrangement comprises a system-on-chip having the one or more than one processor and a phased antenna array as a transmitting and receiving unit (2).

13. Sensor arrangement according to one of the preceding claims, wherein the sensor arrangement comprises a thermopile array for detecting a temperature of an object in the monitoring area (1).

14. Sensor arrangement according to one of the preceding claims, wherein the monitoring area (1) has a size of at least 5m x 5m.

Images