Method for detecting a living being, evaluation unit and vehicle
Patent Information
- Application Number
- EP2024709041
- Authority / Receiving Office
- EP · EP
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2023-03-06
- Filing Date
- 2024-03-01
- Publication Date
- 2026-01-14
AI Technical Summary
Existing methods for detecting small children in vehicles, such as optical cameras, are complex and memory-intensive, making them inefficient for use with sensors like radar or UWB sensors already present in vehicles.
A method that forms a one-dimensional vector from sensor measurements, creates a fluctuating function, and determines the number of mean value crossings to differentiate between types of living beings, such as adults, children, and pets, using existing vehicle sensors like UWB sensors, thereby simplifying evaluation and reducing memory requirements.
This approach enables accurate recognition of living beings with reduced memory usage, allowing for improved detection of small children and differentiation between adults and children based on breathing patterns, enhancing safety by potentially eliminating the need for optical cameras.
Smart Images
Figure EP2024055469_12092024_PF_FP_ABST
Abstract
Description
[0001] Method for detecting a living being, evaluation unit and vehicle
[0002] Description
[0003] The invention relates to a method for detecting a living being and to an evaluation unit for this purpose. Furthermore, the invention relates to an application of the method for detecting a small child in a vehicle and to a vehicle.
[0004] In modern vehicles, it is now advantageous to be able to detect whether a small child has been left behind in the car, as this can pose a significant danger to the small child, particularly in areas with warm climates.
[0005] Methods for detecting small children are generally known, for example using optical cameras.
[0006] It would be advantageous if such a camera were no longer necessary for monitoring the interior and could be replaced by other sensors such as radar sensors or UWB sensors, whereby sensors that are already present in the vehicle are advantageously used.
[0007] The disadvantage of using such sensors is the complex and memory-intensive nature of evaluating the signals provided by the sensors.
[0008] Therefore, the object of the invention is to simplify the evaluation of sensors, in particular to enable it with less memory.
[0009] The object is achieved by a method according to claim 1, an evaluation unit according to claim 9, and a vehicle with such an evaluation unit. The method serves to detect a living being in an area, in particular to detect an adult or a small child in an area. In the method, measured values from at least one sensor are provided to an evaluation unit, wherein the evaluation unit performs at least the following steps:
[0010] - Formation of a one-dimensional vector, wherein the vector comprises temporally arranged measured values of the at least one sensor;
[0011] - Formation of a fluctuating function (F)
[0012] - Formation of an average value of the fluctuating function (F), in particular by averaging temporally recurring measured values (M);
[0013] - Determination of the number of measured values which essentially correspond to the respective mean value or;
[0014] - Determination of the number of mean value passes of a difference between the measured values and the respective mean value of the measured values,
[0015] - Whereby the number of mean value runs and / or the number of measured values that essentially correspond to the respective mean value serves to determine the species of the living being.
[0016] The invention is based on the finding that the signal provided by the respective sensor reflects a movement in the area monitored by the sensor. The signal therefore varies slightly from measurement cycle to measurement cycle. A fluctuating, particularly periodic, movement such as the breathing of a living being in the area accordingly leads to a more or less periodic change in the signal from measurement cycle to measurement cycle.
[0017] Advantageously, the detection of a living being here refers to both the detection of a living being and the classification of the living being. Classification is advantageously a living being such as a small child, an adult, or a pet. An area can be understood as the interior of a vehicle. In particular, the area can be a rear seat bench of a vehicle. Advantageously, the area is defined by the position and / or orientation of at least one sensor. Advantageously, the area is a passenger compartment of a vehicle.
[0018] At a definable point in time within the respective measuring cycle, this can lead to a fluctuating course of the measured values at the definable point in time.
[0019] A fluctuating function is, in particular, a function consisting of a plurality of measured values arranged sequentially over time. The function has a mean value, with the measured values fluctuating around the mean value (of the totality of the measured values). The function is advantageously designed as a mapping of the respective measured value as a function of time. Function values (measured values) are alternately greater than the mean value in one sub-range of the function and smaller than the mean value in a subsequent sub-range of the function.
[0020] The mean value crossings, i.e., the ranges in which at least one measured value essentially corresponds to the mean, can be arranged equidistantly in time or not. With an equidistant arrangement of the mean value crossings, the function fluctuates periodically around the mean.
[0021] The vector advantageously comprises the measured values of the individual measurement cycles arranged sequentially. For example, a measurement cycle comprises 100 measured values, each recorded consecutively every millisecond.
[0022] If the measured values are provided by the respective sensor as complex values, each comprising a magnitude and a phase, the magnitudes can be arranged in a first region of the vector and the phases can be arranged in a second region of the vector. Alternatively, the respective magnitudes of the respective measured value and the respective phase can be arranged alternately in the vector.
[0023] It is advantageous to record the measured values as a recurring function. For example, 100 measured values are recorded within 100 microseconds and the recording is repeated 100 times.
[0024] To facilitate simplified evaluation, the nth measured value is stored in a reduced vector. For example, every 30th measured value can form an entry in a reduced vector. This means that a vector of 10,000 entries is "reduced" to a vector with 100 entries.
[0025] Such a reduced vector can be the basis for the fluctuating function.
[0026] The number of mean value transitions of the function is advantageously determined from the fluctuating function. One possibility is to determine the measured values that correspond to the mean value at each mean value transition.
[0027] Alternatively or additionally, the number of mean value passes can be determined by how many neighboring measured values have a lower value and a higher value than the mean value.
[0028] Alternatively or additionally, the number of mean value passes can also be the number of measured values which at least substantially correspond to the mean value.
[0029] The number of mean value cycles indicates how frequently the fluctuating function changes direction. Rapid movement or a higher breathing rate of the creature can be inferred from a younger creature, such as a toddler.
[0030] Alternatively, or additionally, the temporal distance between the mean value transitions of the fluctuating function can also be used to determine the species of the organism. The shorter the temporal distance, the younger the organism. A short distance therefore indicates a small child, while a long distance is more likely to indicate an adult.
[0031] In addition, a small amplitude of the respective fluctuating function can also be inferred from a smaller or younger organism.
[0032] It is advantageous to detect the movement of the living being's breathing. Breathing (respiration amplitude, respiratory rate) is advantageously represented in the form of a fluctuating function.
[0033] The fluctuating function can be interpreted as a representation of the fluctuating movement, especially periodic movement, of a living being.
[0034] If a fluctuating function is present, an object can be closed that moves fluctuatingly - e.g. through breathing.
[0035] Furthermore, the period of the movement can be inferred based on the number of mean value transitions. For example, the breathing or heart rate of a toddler is higher than that of an adult. Thus, by determining the number of mean value transitions (per unit of time), it is possible to distinguish whether the living being is an adult or a toddler. Advantageously, the vector, the reduced vector, and / or the number of mean value transitions can be provided to a further evaluation algorithm, particularly a neural network.
[0036] The invention explained above enables improved recognition of a living being.
[0037] In an advantageous embodiment of the invention, an amplitude of the fluctuating function is determined between the respective mean value transitions, whereby the amplitude serves to determine the type of living being (adult, pet, or infant). The amplitude can be defined as the largest difference between the respective measured value and the mean value.
[0038] Advantageously, the amplitude of the fluctuating function is determined by determining the difference between the respective function value of the fluctuating function and the mean value halfway between two mean value passages or in the middle between two values which essentially correspond to the mean value.
[0039] The amplitude of the fluctuating function is advantageously a measure of the expression of the movement of a living being, in particular the amplitude of the respective movement, e.g. breathing as an increase in the diameter of the chest.
[0040] For example, the breathing movement of a toddler is not as pronounced as that of an adult. Therefore, if the amplitude of the fluctuating function is high, it is advantageous to assume that an adult is in the area.
[0041] By taking the amplitude of the fluctuating function into account, an improved detection of a living being, and in particular the determination of whether it is an adult or a small child, can be determined with a higher probability than if only the distance or the number of mean value runs were considered.
[0042] In a further advantageous embodiment of the invention, the fluctuating function can be stored as a reduced vector. The reduced vector can be stored, for example, in a memory of an evaluation unit.
[0043] The reduced vector can advantageously be provided to a neural network, whereby the neural network can be made smaller by incorporating the reduced vector than a neural network that can “incorporate” the (entire) vector.
[0044] Advantageously, the fluctuating function contains fewer measured values than the vector from which the fluctuating function is determined. Therefore, the reduced vector can be stored in a smaller memory area than the vector.
[0045] In a further advantageous embodiment of the invention, a decision is made about the type of living being based on the respective number and / or the amplitude, whereby the living being is determined as a toddler if the number of the mean value passes is above a predeterminable number, and if the number is less than the predeterminable number, the living being is determined as an adult
[0046] Determining the species of the organism based on the number of mean value passes and the respective amplitude is advantageous for increasing the certainty in determining the species of the organism.
[0047] In a further advantageous embodiment of the invention, the respective measured value comprises a magnitude and a phase, wherein the respective magnitudes of the measured value are arranged in a first section of the vector and the respective phases are arranged in a second section of the vector. Some sensors provide complex measured values, wherein the respective measured value comprises a magnitude and a phase. Advantageously, the magnitudes and / or the phases can be evaluated together or separately. Advantageously, the vector comprises a first section with magnitudes and a subsequent section with phases. V = (B1 , ... BN,q>1 , ... ,cpN) where B is the respective magnitude and cp is the respective phase of the measured value.
[0048] The division of the vector is also advantageously carried out accordingly for the reduced vector.
[0049] The procedure is advantageously carried out in the same way for both the phases and the magnitudes of the measured values. The process steps described here are advantageously carried out for the magnitudes of the respective measured values and for the respective
[0050] By evaluating amounts and phases separately, a particularly high level of accuracy can be achieved in the detection of living beings.
[0051] In a further advantageous embodiment of the invention, in a first step the recognition of a living being takes place with the aid of the magnitudes of the vector and / or the reduced vector and in a second step the recognition of the living being takes place with the aid of the phases.
[0052] By evaluating amounts and phases separately, a particularly high level of accuracy can be achieved in the detection of living beings.
[0053] In a further advantageous embodiment of the invention, the living being is recognized as a small child if:
[0054] - a predeterminable number of amplitudes remains below a predeterminable amplitude and / or - the number of mean value passes is higher than a predeterminable number (of mean value passes).
[0055] The predeterminable number of amplitudes and the predeterminable number of mean value passes are advantageously determined experimentally.
[0056] By using two experimentally determinable sizes, particularly accurate detection of a small child can be ensured.
[0057] In a further advantageous embodiment of the invention, the vector and the fluctuating function are provided to a neural network, wherein the neural network recognizes the living being on the basis of the fluctuating function and optionally on the basis of the vector and makes a decision as to whether the living being is a small child, an adult, or a pet.
[0058] Advantageously, only the fluctuating function or the reduced vector is provided to the neural network, so that the neural network only needs to be small in size and complexity. This saves the necessary memory in the evaluation unit.
[0059] The evaluation unit is preferably designed as a software module. The evaluation unit is preferably integrated into a vehicle control system.
[0060] The method is advantageously designed as an executable computer program product and can advantageously be executed on a computing unit.
[0061] The evaluation unit has at least one interface for at least one sensor. The interface can be embodied as a software interface. Furthermore, the evaluation unit has a processing unit, wherein the processing unit is designed and provided for executing the method described here. Optionally, the evaluation unit comprises a neural network, wherein the neural network is designed to evaluate the fluctuating function. The neural network is provided for deciding whether a living being is detected and / or for deciding whether the living being is a small child, an adult, or a pet.
[0062] In a further advantageous embodiment of the invention, the at least one sensor is designed as an ultra-wide-band (UWB) sensor.
[0063] The vehicle comprises an evaluation unit as described above and at least one sensor for monitoring the area. The area is advantageously the interior of a passenger compartment or a trunk of the vehicle.
[0064] Instead of a neural network, another learning algorithm can also be used to assist in the recognition and classification of the living being.
[0065] The invention is explained and explained in more detail below with reference to figures. The embodiments shown in the figures are merely exemplary and should not be considered as limiting the invention.
[0066] They show:
[0067] Fig. 1 a vehicle with an exemplary evaluation unit,
[0068] Fig. 2 shows an exemplary process diagram,
[0069] Fig. 3 shows another exemplary process diagram and Fig. 4 shows an example of a fluctuating function.
[0070] Fig. 1 shows a vehicle 10 with an exemplary evaluation unit 7. The vehicle 10 comprises an area 1. The area 1 can be the passenger compartment or the trunk of the vehicle 10. Sensors 3 are used to monitor the area 1. Only one sensor 3 is necessary for the invention. Therefore, the other two sensors 3 are shown in dashed lines. The respective sensor 3 is advantageously designed as an ultra-wideband sensor. The respective sensor 3 provides measured values at respective times t, advantageous to each of the N
[0071] At each point in time, a measured value M1 ,... ,MN.
[0072] A living being Obj is shown in area 1. The sensors 3 are used to detect the living being Obj in area 1. A characteristic of a living being Obj is movement, for example through breathing. This movement is symbolized by the dashed curve on the living being Obj. The sensors 3 are provided and designed to provide measured values M1,..., MN to the evaluation unit 7. The evaluation unit 7 comprises a processing unit 7a and optionally a neural network. The evaluation unit 7 can provide a signal to a signal generator 9 upon detection of a living being Obj, in particular a small child KK (without the simultaneous presence of an adult Erw). The signal generator 9 is used to emit a warning signal if a small child KK has been forgotten in area 1 of the vehicle 10. The signal generator 9 can also be designed with a communication unit to provide a warning signal for a mobile phone.
[0073] Fig. 2 shows an example process diagram. At the top, measured values M are shown as a function of time t. Shown are the measured values M for three measuring cycles M1, MP and MN. At each predeterminable time tx, the measuring points Mx are highlighted as M(tx). The measured values for several measuring cycles P can be stored in a vector V = (M1 (t1),... M1(tN),... ,MP(t1),... ,MP(tN),...., MN(t1),... (MN(tN)).
[0074] The measuring points Mx can also be stored in a reduced vector Vr = (Mx(P=1 ),... , Mx(P=N)) from the respective measuring cycle P. P runs from P=1 to P=N. The measuring points Mx form the fluctuating function F arranged one after the other in time. The fluctuating function F is a function of the respective measuring point Mx from the respective measuring cycle P. The fluctuating function F has a mean value <mx>, where measuring points Mx are arranged around the mean value <mx>fluctuate back and forth. The fluctuating function F is shown at the bottom of the figure.
[0075] The measuring points, which are in the western center <mx>are called mean value crossings ND. Between the mean value crossings ND, the fluctuating function exhibits a minimum or maximum. The difference between the respective minimum / maximum and the mean <mx>is referred to as amplitude. Depending on the deviation of the measured values M from measurement cycle P at the respective time tx, the function fluctuates more or less. With a significant deviation of the measured values from measurement cycle P to measurement cycle P, the number #ND of mean value transitions ND of the fluctuating function increases. Furthermore, the distance d between the respective mean value transitions ND decreases as the number #ND increases.
[0076] Any noise in the measured values can be advantageously eliminated by averaging.
[0077] Fig. 3 shows another exemplary process diagram. The upper part of the figure shows the steps of process unit 7a. Shown is the conversion of vector V into the reduced vector, providing the mean value, the number #ND of mean value passes ND, and the respective amplitude A. The fluctuating function can be provided as a reduced vector Vr.
[0078] Below it is shown that the neural network NN is provided with the vector V (comprising the measured values of the measurement cycles P), the reduced vector V (comprising the fluctuating function F). Furthermore, the mean value <mx>, the respective amplitude A and the number #ND of the mean value passes ND are provided.
[0079] With the help of the neural network, the recognized creature Obj can be recognized as a toddler KK, as a pet HT or as an adult Erw.
[0080] Fig. 4 shows an example of a fluctuating function F. In this example, the measured values M provided by the respective sensor 3 are provided in complex form (as complex numbers). The fluctuating function F can have a first section S1 and a second section S2. The first section S1 comprises the absolute values B of the respective measuring points Mx. The second section S2 comprises the phases cp of the respective measuring points Mx (each in a temporal arrangement). The evaluation of the fluctuating function F is carried out, as shown here, once using the phases cp and once using the absolute values B. In other words, once with the first section S1 of the (reduced) vector and once with the second section S2 of the (reduced) vector.
[0081] It is visible that the fluctuating function F can have a different course for the amounts B and the phases cp.
[0082] Such a fluctuating function F can advantageously be determined by a vector V, where the vector V has a section with amounts B (of the respective measured value M) and a section with phases cp (of the respective measured value M)
[0083] Instead of the fluctuating function F, the respective measured values M can also be displayed accordingly.
[0084] 1 area
[0085] 3 Sensor (especially a UWB sensor)
[0086] Obj living beings
[0087] 7 Evaluation unit
[0088] 7a Process unit
[0089] 10 vehicles
[0090] NN Neural Network
[0091] V Vector
[0092] M1 ,MN measured values (provided by the sensor)
[0093] F Fluctuating function
[0094] Mx measuring point
[0095] ND mean value passage
[0096] A Amplitude (of F) t Time tx Time
[0097] Vr reduced vector d distance (between each two mean value passes)
[0098] S1 , S2 First, second section
[0099] <mx>Mean value (of the measurement points) or the fluctuating function
[0100] B Amount (of the measuring point or the measured value) cp Phase
[0101] 1 , ... m No. of the measuring cycle
[0102] KK Toddler
[0103] Adult
[0104] HT Pet
[0105] P measuring cycle
[0106] #ND Number (of mean value passes) In summary, the invention relates to a method for detecting a living being Obj, an evaluation unit 7 for this purpose and a vehicle 10 with such an evaluation unit 7. For this purpose, at least one sensor 3 records measured values M1,...,MN (P=1, ... N) in several measuring cycles P, which can be arranged consecutively in a vector V, wherein the vector V is provided to an evaluation unit 7. The vector V comprises the recorded measured values M in temporal succession. At a definable time tx, a measuring point Mx is determined from the measured values M. The measuring points form a fluctuating function F. The fluctuating function is averaged and a mean value <mx>is provided. In other words, an average <mx>the respective n-th measured values Mx are calculated.
[0107] Advantageously, the respective measuring point Mx is selected such that the respective measuring points result in a more or less periodic, i.e., fluctuating function F with an amplitude A and a number #ND of mean value transitions ND. Based on the amplitude A and the number #ND, a living being Obj is recognized, in particular whether the living being Obj is an adult Erw, a pet HT, or a small child KK.< / mx> < / mx> < / mx> < / mx> < / mx> < / mx> < / mx> < / mx>
Claims
Method for detecting a living being, evaluation unit and vehicle Patent claims 1. Method for detecting a living being (Obj) in an area (1), in particular an adult (Erw) or a small child (KK) in an area (1), in particular in a vehicle (10), wherein measured values (M1,...,MN) from at least one sensor (3) are provided to an evaluation unit (7) during a measuring cycle (m), wherein the evaluation unit (7) carries out at least the following steps: - Formation of a vector (V), wherein the vector (V) comprises temporally arranged measured values (M1,... MN) of the at least one sensor (3); - Formation of a fluctuating function (F) from measuring points (MX), where each measuring point (Mx) is a measured value (M1,...,MN) at a predeterminable time (tx), - Formation of an average ( <mx>) of the fluctuating function (F), in particular by averaging the measuring points (Mx); - Determination of the number of measuring points (Mx), which essentially correspond to the respective mean value ( <mx>) correspond or; - Determination of a number (#ND) of mean value passes (ND) of a difference between the measuring points (Mx) and the respective mean value ( <mx>) of the measuring points (Mx); - Where the number (#ND) of mean value runs (ND) and / or the number of measuring points (Mx) that essentially correspond to the respective mean value ( <mx>) is used to determine the type of living being (Obj).
2. Method according to claim 1, wherein an amplitude (A) of the fluctuating function (F) is determined between the respective mean value passages (ND), wherein the amplitude (A) serves to determine the type of living being (5).
3. Method according to one of the preceding claims, wherein the fluctuating function (F) can be stored in a reduced vector (Vr).
4. Method according to one of the preceding claims, wherein a decision on the type of living being (Obj) is made on the basis of the respective number (#ND) of the mean value passages and / or the amplitude (A), wherein the living being (Obj) is determined as a small child (KK) if the number (#ND) of the mean value passages (ND) is greater than a predeterminable number (#ND), and if the number (#ND) is less than the predeterminable number (#ND), the living being is determined as an adult (Erw).
5. Method according to one of the preceding claims, wherein the respective measured value (M1,..., MN) comprises in each case an amount (B) and in each case a phase (cp), wherein the respective amount (B) of the measured value (M1,..., MN) is arranged in a first section (S1) of the vector (V) and the respective phases (cp) are arranged in a second section (S2) of the vector (V).
6. Method according to the preceding claim, wherein in a first step the recognition of the living being (Obj) is carried out with the aid of the magnitudes (B) of a vector (V) and in a second step the recognition is carried out with the aid of the phases (cp).
7. Method according to one of the preceding claims, wherein the living being (Obj) is recognized as a small child (KK) if: - a predeterminable number of amplitudes (A) remains below a predeterminable amplitude (A*) and / or - the number (#ND) of mean value passes (ND) is higher than a specified number.
8. Method according to one of the preceding claims, wherein the vector (V) and the fluctuating function (F) are provided to a neural network (NN), wherein the neural network (NN) recognizes the living being (5) on the basis of the fluctuating function (F) and the vector (V) and makes a decision as to whether the living being (Obj) is a small child (KK), a pet (HT) or an adult (Erw).
9. Evaluation unit (7), having at least one interface for at least one sensor (3), wherein the evaluation unit (7) comprises a process unit (7a) and optionally a neural network (NN), wherein the evaluation unit (7) is designed to carry out a method according to one of the preceding claims.
10. Evaluation unit according to the preceding claim, wherein the sensor (3) is an ultra-wide-band sensor.
11. Vehicle (10), comprising an evaluation unit (7) according to one of claims 9 or 10, and at least one sensor (3) for monitoring the area (1).< / mx> < / mx> < / mx> < / mx>