DE102014019760A1 - Method and device for determining the time course of the respiratory depth - Google Patents

Abstract

The invention relates to a method and a device for determining the time course of the breathing depth (T) of a, in particular sleeping, person. There are continuously created at individual recording times height profiles (H) of the person (1). Height profiles from adjacent recording times are combined into segments. The area indicating the abdominal or chest area of the person depending on the respective reference point or reference area (21) is selected as the viewing area (22). For each height profile (H) within the segment (S1,..., Sq), the mean value of the distances of the points of the height profile (H) within the viewing area (22) to a reference point or reference object is determined separately. For the segment (S1,..., Sq), a signal (si) is determined, to which the mean value determined for this height profile (H) is assigned to the respective recording time (t1,..., Tp) of the height profile (H). Based on the determined signal (si), in particular based on its signal amplitude, one or more values characterizing the temporal course of the breathing depth (T) are determined

Description

The invention relates to a method for determining the time course of the breathing depth of a, in particular sleeping, person. Furthermore, the invention relates to a device for determining the time course of the respiratory depth. Finally, the invention relates to a data carrier with a program for determining the time course of the breathing depth.

There are continuously created at individual recording times height profiles of the person. Height profiles from adjacent recording times are grouped into segments. The area indicating the abdominal or chest area of the person depending on the respective reference point or reference area is selected as the viewing area.

For each height profile within the segment, the mean value of the distances of the points of the height profile located within the viewing area to a reference point or reference object is determined separately in each case. For the segment, a signal is determined, which is assigned to the respective recording time of the height profile of the determined for this height profile average.

Based on the detected signal, in particular based on its signal amplitude, one or more values characterizing the temporal course of the breathing depth are determined

From the prior art, a variety of methods are known, with which the breathability of a person, especially during sleep, is recorded. In these known methods, the person to be examined is in each case directly connected to different sensors which can disturb the person during sleep. Also, the cables and wires sometimes used for immediate evaluation of the sensor signals can disturb the sleep of the person concerned.

The object of the invention is therefore to provide a method and a device which is not very disturbing for the person to be examined and in particular determines the respiration without contact.

The invention solves this problem in the aforementioned method with the features of claim 1. The invention solves this problem in the aforementioned device with the features of claim 18th

• a) wobei im Höhenprofil eine Anzahl von zumindest zwei Punkten im Raum festgelegt ist, die auf der Oberfläche der Person oder auf der Oberfläche eines auf oder neben der Person befindlichen Gegenstands liegen,
• b) wobei für jeden der Aufnahmezeitpunkte das jeweilige Höhenprofil in einer Datenstruktur abgespeichert und zur Verfügung gehalten wird,
• c) wobei eine Anzahl von zu aufeinander folgenden Aufnahmezeitpunkten, insbesondere innerhalb eines Zeitbereichs von 3 bis 20 Sekunden, aufgenommenen Höhenprofilen zu einem Segment zusammengefasst werden,
• d) wobei ein Bereich, der den Bauch- oder Brustbereich der Person in Abhängigkeit vom jeweiligen Referenzpunkt oder Referenzbereich angibt als Betrachtungsbereich ausgewählt wird,
• e) wobei für jedes Höhenprofil innerhalb des Segments jeweils gesondert der Mittelwert der Abstände der innerhalb des Betrachtungsbereichs befindlichen Punkte des Höhenprofils zu einem Referenzpunkt oder Referenzgegenstand ermittelt wird, und für das Segment ein Signal ermittelt wird, dem zum jeweiligen Aufnahmezeitpunkt des Höhenprofils der für dieses Höhenprofil ermittelte Mittelwert zugeordnet wird, und
• f) wobei basierend auf dem ermittelten Signal, insbesondere basierend auf dessen Signalamplitude, ein oder mehrere den zeitlichen Verlauf der Atemtiefe charakterisierende Werte ermittelt wird oder werden.

In a method for determining the temporal progression of the breathing depth of a person, in particular a sleeping person, the invention provides that a height profile of the person is continuously produced with a detector unit directed at the person continuously at successive recording times,

• a) wherein the height profile defines a number of at least two points in the space lying on the surface of the person or on the surface of an object located on or next to the person,
• b) wherein for each of the recording times the respective height profile is stored in a data structure and made available,
• c) wherein a number of height profiles recorded at successive recording times, in particular within a time range of 3 to 20 seconds, are combined to form a segment,
• d) wherein an area indicating the abdominal or chest area of the person as a function of the respective reference point or reference area is selected as the viewing area,
• e) wherein for each height profile within the segment each separately the average of the distances of the points located within the viewing range of the height profile is determined to a reference point or reference object, and for the segment, a signal is determined to the respective recording time of the height profile of the height profile determined mean value is assigned, and
• f) wherein based on the detected signal, in particular based on the signal amplitude, one or more of the temporal course of the breathing depth characterizing values is or will be determined.

Through these advantageous measures, a non-contact detection of the breathing of a person is made possible, which does not disturb the person to be examined while sleeping. In particular, it is not necessary to attach sensors to the body of the person to be examined.

• a) wobei das Höhenprofil eine Anzahl von zumindest zwei Abstandswerten zur Festlegung jeweils eines Punktes im Raum aufweist, wobei die einzelnen Abstandswerte jeweils den Abstand des Schnittpunktes eines relativ zur Detektoreinheit vorab festgelegten, insbesondere von der Detektoreinheit ausgehenden, Strahls mit der Oberfläche der Person oder der Oberfläche eines auf der oder neben der Person befindlichen Gegenstands von einem Referenzpunkt oder einer Referenzebene angeben
• b) wobei für jeden der Aufnahmezeitpunkte jeweils eine Datenstruktur erstellt wird, die das jeweilige Höhenprofil enthält, wobei sämtliche so erstellte Datenstrukturen jeweils dieselbe Größe aufweisen und jeweils mit Speicherpositionen für die einzelnen Abstandswerte des Höhenprofils aufweisen,
• c) wobei eine Anzahl von zu aufeinander folgenden Aufnahmezeitpunkten, insbesondere innerhalb eines Zeitbereichs von 3 bis 20 Sekunden, aufgenommenen Höhenprofilen zu einem Segment zusammengefasst wird,
• d) wobei eine Anzahl von Speicherpositionen der Datenstruktur, in denen Abstandswerte abgespeichert sind, die den Abstand des Bauch- oder Brustbereichs der Person in Abhängigkeit vom jeweiligen Referenzpunkt oder Referenzbereich angeben, als Betrachtungsbereich (ausgewählt wird,
• e) wobei für jedes Höhenprofil innerhalb eines Segments jeweils gesondert der Mittelwert der innerhalb des Betrachtungsbereichs befindlichen Abstandswerte ermittelt wird, und für das Segment ein Signal ermittelt wird, dem zum jeweiligen Aufnahmezeitpunkt des Höhenprofils der für dieses Höhenprofil ermittelte Mittelwert zugeordnet wird, und
• f) wobei basierend auf dem ermittelten Signal, insbesondere basierend auf dessen Signalamplitude, ein oder mehrere den zeitlichen Verlauf der Atemtiefe charakterisierende Werte ermittelt wird oder werden.

An advantageous and numerically simple embodiment of the invention provides that for determining the time course of the breathing depth of a, in particular sleeping, person with a person on the person Directional detector unit is continuously created at successive recording times in each case a height profile of the person,

• a) wherein the height profile has a number of at least two distance values for defining a point in space, the individual distance values in each case the distance of the intersection of a predetermined relative to the detector unit, in particular emanating from the detector unit, the beam with the surface of the person or Indicate the surface of an object on or next to the person from a reference point or a reference plane
• b) wherein for each of the recording times in each case a data structure is created which contains the respective height profile, wherein all the data structures thus created each have the same size and in each case have memory positions for the individual distance values of the height profile,
• c) wherein a number of height profiles taken at successive recording times, in particular within a time range of 3 to 20 seconds, is combined to form a segment,
• d) wherein a number of memory positions of the data structure in which distance values are stored, which indicate the distance of the abdominal or chest region of the person as a function of the respective reference point or reference region, is selected as the viewing region (FIG.
• e) wherein, for each height profile within a segment, the mean value of the distance values located within the viewing area is determined separately, and a signal is determined for the segment to which the average value determined for this height profile is assigned at the respective time of recording of the height profile, and
• f) wherein based on the detected signal, in particular based on the signal amplitude, one or more of the temporal course of the breathing depth characterizing values is or will be determined.

An alternative advantageous and numerically precise embodiment of the invention, which allows the use of height profiles in the form of general point clouds, provides that the height profile describes a point cloud with a number of at least two points in space on the surface of the person or on the Surface of an object located on or next to the person.

An advantageous extraction of the respiration depth from the signal is achieved by extracting at least one maximum and at least one minimum from the signal for characterizing the respiratory depth and at least one difference between the maximum and the minimum as the value characterizing the respiration depth for the segment assigned time range is used.

An alternative advantageous extraction of the depth of respiration from the signal is achieved by subjecting the signal to a spectral transformation, in particular Fourier transformation or cosine transformation or wavelet transformation, and within a given frequency band, in particular from 0.1 Hz to 1 Hz, after that Spectral component with the highest signal levels is sought and the signal energy of this spectral component is used to characterize the breathing depth in this segment.

• a) Signalanteile mit einer Frequenz von mehr als 0,5 Hz unterdrückt werden, und/oder
• b) Gleichanteile unterdrückt werden.

For low-noise extraction of respiratory movements, from which ultimately a particularly good extraction of the respiratory depth is possible, it can be provided that the signal assigned to a segment is subjected to noise filtering after its creation, in particular before the determination of the value characterizing the respiration depth, wherein in particular

• a) signal components are suppressed with a frequency of more than 0.5 Hz, and / or
• b) Gleichanteile be suppressed.

The course of the breathing depth over time can advantageously be determined by separately determining the breathing depth for a number of overlapping or non-overlapping segments.

An advantageous adaptive adaptation of the viewing area, which can be performed regularly with little effort, provides that the viewing area for each segment is determined separately, in particular on the basis of the viewing area determined for the respective preceding segment.

• – ein Anzahl von in Zeilen und Spalten rasterförmig angeordneten Positionen vorgegeben wird, an denen jeweils die Abstandswerte des Höhenprofils bestimmt werden
• – die Matrixdatenstruktur ein Raster von derselben Größe und Struktur aufweist, und
• – die Matrixdatenstruktur erstellt wird, indem die an den jeweiligen Positionen aufgenommenen Abstandswerte an den den Positionen im Raster entsprechenden Speicherpositionen in der Matrixdatenstruktur abgespeichert und zur Verfügung gehalten werden.

Particularly meaningful height profiles can be created by: characterizing the height profile by a two-dimensional matrix data structure comprising a number of rows and columns,

• - Given a number of arranged in rows and columns grid-like positions, at each of which the distance values of the height profile are determined
• The matrix data structure has a grid of the same size and structure, and
• - The matrix data structure is created by the stored at the respective positions distance values are stored at the positions corresponding to the grid positions in the matrix data structure and kept available.

In order to enable a faster processing of the data, to reduce the required storage space and to reduce the occurring noise, it may be provided that the matrix data structure is dimensioned after its creation by a reduced matrix data structure, covering for rectangular, in particular the entire matrix data structure , and non-overlapping image areas of the same size in the matrix data structure in each case a mean distance value is determined and this average distance value is assigned to the corresponding pixel with respect to its position of the reduced matrix data structure.

• a) im Höhenprofil, insbesondere in der Matrixdatenstruktur oder in der Punktwolke, eine Anzahl von möglichen Betrachtungsbereichen vorab vorgegeben werden,
• b) für das jeweilige Segment die jeweilige Atemtiefe unter Zugrundelegung jedes der möglichen Betrachtungsbereiche ermittelt wird, und
• c) derjenige vorgegebene möglichen Betrachtungsbereich als Betrachtungsbereich herangezogen wird, für den die ermittelte Atemtiefe am größten ist.

An advantageous method for determining the viewing area provides that the viewing area is set by

• a) a number of possible viewing ranges are predefined in the height profile, in particular in the matrix data structure or in the point cloud,
• b) for each segment the respective breathing depth is determined on the basis of each of the possible viewing areas, and
• c) that predetermined possible viewing area is used as the viewing area for which the determined breathing depth is greatest.

• a) im Höhenprofil, insbesondere in der Matrixdatenstruktur oder in der Punktwolke, mittels Objekterkennung nach Bereichen gesucht wird oder dass vorab Bereiche ausgewählt werden, die einem menschlichen Kopf und Torso entsprechen, und
• b) der den Torso abbildende Bereich des Höhenprofils, insbesondere der Matrixdatenstruktur oder der Punktwolke, oder ein Teilbereich dieses Bereichs als Betrachtungsbereich ausgewählt wird, wobei insbesondere
• – zur Detektion des Verlaufs oder der Atemtiefe der Brustatmung der dem Kopf nahe Teilbereich dieses Bereichs als Betrachtungsbereich ausgewählt wird und/oder
• – zur Detektion des Verlaufs oder der Atemtiefe der Bauchatmung der dem Kopf ferne Teilbereich dieses Bereichs als Betrachtungsbereich ausgewählt wird.

A second advantageous method for determining the viewing area, which takes particular account of anatomical conditions, provides that the viewing area is set by

• a) in the height profile, in particular in the matrix data structure or in the point cloud, by means of object recognition is searched for areas or that areas are selected in advance, which correspond to a human head and torso, and
• b) the torso-imaging region of the height profile, in particular the matrix data structure or the point cloud, or a subregion of this region is selected as the viewing region, wherein in particular
• - For detecting the course or the breathing depth of the chest breathing of the head near portion of this area is selected as a viewing area and / or
• - For detecting the course or the breathing depth of the abdominal breathing of the head remote portion of this area is selected as the viewing area.

A rapid adaptation of the viewing area can be achieved by adaptively recalculating the viewing area for each segment to be examined or for individual height profiles of a segment by means of object recognition to the height profiles present in the segment starting from the position of the selected viewing area in the respective preceding segment.

• a) für sämtliche Punkte des Höhenprofils, insbesondere für alle Einträge der Matrixdatenstruktur oder der Punktwolke, separat die Varianz über das jeweilige Segment bestimmt wird, und
• b) ein Bereich oder mehrere Bereiche mit jeweils zusammenhängenden Punkten des Höhenprofils, deren jeweilige Varianz oberhalb eines unteren Schwellenwerts oder innerhalb eines vorgegebenen Intervalls liegen, als Betrachtungsbereich ausgewählt wird oder werden, insbesondere
• – indem zusammenhängende Bereiche von Punkten als Betrachtungsbereich ausgewählt werden, deren Größe eine vorgegebene Anzahl von Punkten übersteigt
• – wobei insbesondere ein Bereich dann als zusammenhängend angesehen wird, wenn jeder Punkt des Höhenprofils ausgehend von einem Punkt innerhalb dieses Bereichs über jeweils benachbarte Pixel erreichbar ist,
• – wobei vorzugsweise Punkte des Höhenprofils dann als benachbart angesehen werden,
• – wenn sie durch benachbarte Einträge der Matrix-Datenstruktur festgelegt sind oder
• – wenn sie oder ihre Projektion auf eine horizontale Ebene im Raum einen einen Schwellenwert unterschreitenden Abstand voneinander aufweisen.

A third advantageous method for determining the viewing area, which is numerically stable and efficiently executable, provides that the viewing area is set by

• a) for all points of the height profile, in particular for all entries of the matrix data structure or the point cloud, separately the variance over the respective segment is determined, and
• b) an area or a plurality of areas, each having contiguous points of the height profile whose respective variance is above a lower threshold or within a predetermined interval, is or are selected as the viewing area, in particular
• - by selecting contiguous areas of points as a viewing area whose size exceeds a given number of points
• In particular, a region is considered to be contiguous if each point of the height profile can be reached, starting from a point within this region, via respectively neighboring pixels,
• Preferably points of the height profile are then considered to be adjacent,
• If they are defined by adjacent entries of the matrix data structure or
• - If they or their projection on a horizontal plane in space have a threshold below the threshold distance from each other.

In order to be able to detect situations in which the person to be examined moves and to be able to reject the erroneous measured values recorded in these situations, it can be provided that the viewing area is determined separately in two or more consecutive segments and a height profile is discarded and one Error message is issued when the viewing area, in particular its size and / or its center of gravity, relative to the respective preceding segment shifts by a predetermined threshold.

• a) gleichzeitig parallel zur Aufnahme des Abstands der Person von der Detektoreinheit der von der Person abgegebene Schall ermittelt und in Form eines Audio-Signals mit einer Anzahl von Audio-Abtastwerten zur Verfügung gehalten wird,
• b) das Audio-Signal in eine Anzahl von Audio-Segmente, insbesondere mit einer Länge von 100 ms bis 1 Sekunde, unterteilt wird und für jedes der Audio-Segmente untersucht wird, ob in diesem menschliche Atemgeräusche oder andere Geräusche enthalten sind, und für jedes Audio-Segment jeweils ein Klassifikationsergebnis zur Verfügung gehalten wird,
• c) zeitgleich aufgenommene Audio-Segmente und Segmente einander zugeordnet werden, und
• d) nach Audio-Segmenten oder Segmenten mit geringer Atemtiefe oder mit fehlenden Atemgeräuschen gesucht wird und die einem solchen Audio-Segment oder Segment zugeordneten Segmente oder Audio-Segmente ebenfalls auf das Vorliegen geringer Atemtiefe oder fehlender Atemgeräuschen untersucht werden, und
• e) für den Fall, dass in einander zugeordneten Segmenten und Audio-Segmenten jeweils geringe Atemtiefe und fehlender Atemgeräusche detektiert wurden, das Fehlen der Atmung der Person festgestellt wird.

In addition, an acoustic control may be provided for improved detection of respiration. It is provided that

• a) simultaneously detected parallel to the distance of the person from the detector unit of the sound emitted by the person and is kept in the form of an audio signal with a number of audio samples,
• b) the audio signal is divided into a number of audio segments, in particular with a length of 100 ms to 1 second, and is examined for each of the audio segments, whether this human breath sounds or other sounds are included, and for each audio segment is given a classification result,
• c) at the same time recorded audio segments and segments are assigned to each other, and
• d) searching for audio segments or segments with low breathing depth or with missing breath sounds and the segments or audio segments assigned to such an audio segment or segment are also examined for the presence of low breathing depth or absence of breathing sounds, and
• e) in the event that in each case associated segments and audio segments each low breathing depth and lack of breath sounds were detected, the absence of respiration of the person is detected.

• a) ein Klassifikationssignal erstellt wird, das für jedes Audio-Segment jeweils zumindest ein Klassifikationsergebnis enthält, das die Art und Stärke des jeweiligen Geräuschs im Zeitbereich des jeweiligen Audio-Segment angibt,
• b) dass jeweils eine Anzahl von aufeinander folgenden Audio-Segmente zu weiteren Audio-Segmenten, insbesondere mit einer Dauer von 5 bis 10 Sekunden zusammengefasst werden, die vorzugsweise dieselben Zeitbereiche umfassen wie die Segmente,
• c) über die innerhalb eines weiteren Audio-Segments enthaltenen Klassifikationsergebnisse eine Mittelung vorgenommen wird und dem weiteren Audio-Segment der jeweilige Mittelwert zugewiesen wird,
• d) ein weiteres Klassifikationssignal durch Interpolation der Mittelwerte der weiteren Audio-Segmente erstellt wird,
• e) im weiteren Klassifikationssignal sowie im Atemtiefesignal nach Zeitpunkten gesucht wird, zu denen in den beiden Signalen starke Änderungen auftreten,
• f) die so erkannten Zeitpunkte werden als umso relevanter eingestuft werden, je stärker die Änderung des jeweiligen Signals zum jeweiligen Zeitpunkt ist,
• g) diesen Zeitpunkten jeweils ein diesbezüglicher Relevanzwert zugeordnet wird, und
• h) jene Zeitpunkte als Startpunkte oder Endpunkte eines Atemstillstandes detektiert werden, für welche der Betrag des Relevanzwerts über dem Betrag eines Schwellwert liegt, wobei insbesondere
• – der Schwellenwert gebildet wird, indem über einen Zeitbereich, insbesondere vor und/oder nach dem Vergleichszeitpunkt mit dem Referenzmaß ein Mittelwert des Referenzmaßes gebildet wird und der Schwellenwert im Bereich zwischen 120% und 200% dieses Mittelwerts festgelegt wird.

A particularly preferred embodiment of the invention, which enables detection of the beginning and end of respiratory misfires, provides that

• a) a classification signal is generated which contains for each audio segment at least one classification result which indicates the type and strength of the respective noise in the time domain of the respective audio segment,
• b) that in each case a number of consecutive audio segments are combined to form further audio segments, in particular with a duration of 5 to 10 seconds, which preferably comprise the same time ranges as the segments,
• c) an averaging is performed on the classification results contained within a further audio segment and the respective average value is assigned to the further audio segment,
• d) a further classification signal is created by interpolation of the mean values of the further audio segments,
• e) is searched in the further classification signal as well as in the respiratory depression signal according to times at which strong changes occur in the two signals,
• (f) the times thus identified will be considered to be the more relevant the more the change in the signal at that time is,
• g) a relevant relevance value is assigned to each of these times, and
• h) those times are detected as starting points or end points of a respiratory arrest, for which the amount of the relevance value is above the magnitude of a threshold, in particular
• - The threshold value is formed by a mean value of the reference measure is formed over a time range, in particular before and / or after the comparison time with the reference measure and the threshold is set in the range between 120% and 200% of this average.

A method for carrying out a method according to the invention can advantageously be stored on a data carrier.

• – eine auf eine Schlafgelegenheit, insbesondere ein Bett ausrichtbare Detektoreinheit, die laufend zu aufeinander folgenden Aufnahmezeitpunkten jeweils ein Höhenprofil der Person erstellt, wobei im Höhenprofil eine Anzahl von zumindest zwei Punkten im Raum festgelegt ist, die auf der Oberfläche der Person oder auf der Oberfläche eines auf oder neben der Person befindlichen Gegenstands liegen, sowie
• – eine Verarbeitungseinheit, die
• a) für jeden der Aufnahmezeitpunkte das jeweilige Höhenprofil in einer Datenstruktur abgespeichert und zur Verfügung hält,
• b) eine Anzahl von zu aufeinander folgenden Aufnahmezeitpunkten, insbesondere innerhalb eines Zeitbereichs von 3 bis 20 Sekunden, aufgenommenen Höhenprofilen zu einem Segment zusammenfasst,
• c) einen Bereich, der den Bauch- oder Brustbereich der Person in Abhängigkeit vom jeweiligen Referenzpunkt oder Referenzbereich angibt als Betrachtungsbereich auswählt,
• d) für jedes Höhenprofil innerhalb des Segments jeweils gesondert den Mittelwert der Abstände der innerhalb des Betrachtungsbereichs befindlichen Punkte des Höhenprofils zu einem Referenzpunkt oder Referenzgegenstand ermittelt, und für das Segment ein Signal ermittelt wird, dem zum jeweiligen Aufnahmezeitpunkt des Höhenprofils der für dieses Höhenprofil ermittelte Mittelwert zugeordnet ist, und
• e) insbesondere anschließend basierend auf diesem Signal oder dessen Signalamplitude einen die Atemtiefe charakterisierender Wert ermittelt.

The invention further relates to a device for determining the temporal course of the breathing depth of a, in particular sleeping, person comprising

• A detector unit which can be oriented towards a sleeping accommodation, in particular a bed, which continuously generates a height profile of the person at successive recording times, wherein in the height profile a number of at least two points in the room are defined on the surface of the person or on the surface of a person lying on or next to the person, as well as
• A processing unit, the
• a) stores and makes available the respective height profile for each of the recording times in a data structure,
• b) a number of height profiles recorded at successive recording times, in particular within a time range of 3 to 20 seconds, into a segment,
• c) selects an area which indicates the abdominal or chest area of the person as a function of the respective reference point or reference area as the viewing area,
• d) for each height profile within the segment separately determines the average of the distances of the points located within the viewing range of the height profile to a reference point or reference object, and for the segment a signal is determined to the respective recording time of the height profile of the determined for this height profile mean is assigned, and
• e) determines, in particular subsequently based on this signal or its signal amplitude, a value characterizing the breathing depth.

Through these advantageous measures, a non-contact detection of the breathing of a person is made possible, which does not disturb the person to be examined while sleeping. In particular, it is not necessary to attach sensors to the body of the person to be examined.

• – eine auf eine Schlafgelegenheit, insbesondere ein Bett, ausrichtbare Detektoreinheit, die laufend zu aufeinander folgenden Aufnahmezeitpunkten jeweils ein Höhenprofil der Person erstellt, wobei das Höhenprofil eine Anzahl von zumindest zwei Abstandswerten zur Festlegung jeweils eines Punktes im Raum aufweist, wobei die einzelnen Abstandswerte jeweils den Abstand des Schnittpunktes eines relativ zur Detektoreinheit vorab festgelegten, insbesondere von der Detektoreinheit ausgehenden, Strahls mit der Oberfläche der Person oder der Oberfläche eines auf der oder neben der Person befindlichen Gegenstands von einem Referenzpunkt oder einer Referenzebene angeben sowie
• – eine Verarbeitungseinheit, die
• a) für jeden der Aufnahmezeitpunkte jeweils eine Datenstruktur erstellt, die das jeweilige Höhenprofil enthält, wobei sämtliche so erstellte Datenstrukturen jeweils dieselbe Größe aufweisen und jeweils mit Speicherpositionen für die einzelnen Abstandswerte des Höhenprofils aufweisen,
• b) eine Anzahl von zu aufeinander folgenden Aufnahmezeitpunkten, insbesondere innerhalb eines Zeitbereichs von 3 bis 20 Sekunden, aufgenommenen Höhenprofilen zu einem Segment zusammenfasst,
• c) eine Anzahl von Speicherpositionen der Datenstruktur, in denen Abstandswerte abgespeichert sind, die den Abstand des Bauch- oder Brustbereichs der Person in Abhängigkeit vom jeweiligen Referenzpunkt oder Referenzbereich angeben, als Betrachtungsbereich auswählt,
• d) für jedes Höhenprofil innerhalb eines Segments jeweils gesondert den Mittelwert der innerhalb des Betrachtungsbereichs befindlichen Abstandswerte ermittelt, und für das Segment ein Signal ermittelt, dem zum jeweiligen Aufnahmezeitpunkt des Höhenprofils der für dieses Höhenprofil ermittelte Mittelwert zugeordnet ist,
• e) insbesondere anschließend basierend auf diesem Signal oder dessen Signalamplitude einen die Atemtiefe charakterisierender Wert ermittelt.

An advantageous and numerically simple embodiment of the invention provides a device for determining the time course of the breathing depth of a, in particular sleeping, person comprising

• A detector unit which can be oriented to a sleeping accommodation, in particular a bed, which continuously generates a height profile of the person at successive recording times, wherein the height profile has a number of at least two distance values for fixing one point each in space, the individual distance values corresponding to each Indicate the distance of the point of intersection of a beam predefined relative to the detector unit, in particular emanating from the detector unit, with the surface of the person or the surface of an object located on or next to the person from a reference point or reference plane;
• A processing unit, the
• a) creates a data structure for each of the recording times, which contains the respective height profile, wherein all the data structures thus created each have the same size and in each case have memory positions for the individual distance values of the height profile,
• b) a number of height profiles recorded at successive recording times, in particular within a time range of 3 to 20 seconds, into a segment,
• c) a number of memory positions of the data structure in which distance values are stored, which indicate the distance of the abdominal or chest region of the person as a function of the respective reference point or reference region, as the viewing region,
• d) determines separately for each height profile within a segment the average value of the distance values located within the viewing range, and determines for the segment a signal to which the average value determined for this height profile is assigned at the respective recording time of the height profile,
• e) determines, in particular subsequently based on this signal or its signal amplitude, a value characterizing the breathing depth.

An alternative advantageous and numerically precise embodiment of the invention, which allows the use of height profiles in the form of general point clouds, provides that the detector unit creates the height profiles in the form of point clouds with a number of at least two points in space, the points on the Surface of the person or on the surface of an object located on or next to the person.

In a device for determining the time course of the breathing depth of a person, in particular a sleeping person, the invention provides that the processing unit for characterizing the breathing depth in a segment extracts at least one maximum and at least one minimum from the signal and the at least one difference between the maximum and the minimum as the breathing depth characterizing value for the segment associated with the time period available and possibly zoom.

A numerically simple filtering of the signal which largely suppresses disturbing noise provides that the processing unit subjects the signal to a spectral transformation, in particular Fourier transformation, cosine transformation or wavelet transformation, and within a predetermined frequency band, in particular from 0.1 Hz to 1 Hz , searches for the spectral component with the highest signal level and uses the signal energy of this spectral component to characterize the breathing depth in this segment.

• a) Signalanteile mit einer Frequenz von mehr als 0,5 Hz unterdrückt, und/oder
• b) Gleichanteile unterdrückt.

For low-noise extraction of respiratory movements, from which ultimately a particularly good extraction of the respiratory depth is possible, it can be provided that the processing unit subjects the signal assigned to a segment to noise filtering after its creation, in particular before the determination of the value characterizing the respiration depth in particular

• a) suppresses signal components with a frequency of more than 0.5 Hz, and / or
• b) Equal shares suppressed.

The course of the breathing depth over time can advantageously be determined by the processing unit separately determining the breathing depth for a number of overlapping or non-overlapping segments.

An advantageous adaptive adaptation of the viewing area, which can be performed regularly with little effort, provides that the processing unit determines the viewing area for each segment separately, in particular on the basis of the viewing area determined for the respective preceding segment.

• – das Höhenprofil durch eine zweidimensionale Matrixdatenstruktur umfassend eine Anzahl von Zeilen und Spalten charakterisiert,
• – ein Anzahl von in Zeilen und Spalten rasterförmig angeordneten Positionen vorgegeben wird, an denen jeweils die Abstandswerte des Höhenprofils bestimmt
• – die Matrixdatenstruktur ein Raster von derselben Größe und Struktur aufweist, und
• – die Matrixdatenstruktur erstellt, indem sie die an den jeweiligen Positionen aufgenommenen Abstandswerte an den den Positionen im Raster entsprechenden Speicherpositionen in der Matrixdatenstruktur abspeichert und zur Verfügung hält.

Particularly meaningful and numerically easy to handle height profiles can be created by the processing unit

• Characterized the height profile by a two-dimensional matrix data structure comprising a number of rows and columns,
• - Given a number of arranged in rows and columns grid-like positions, at which each determines the distance values of the height profile
• The matrix data structure has a grid of the same size and structure, and
• - Creates the matrix data structure by storing the stored at the respective positions distance values at the positions corresponding to the grid positions in the matrix data structure and keeps available.

In order to enable a faster processing of the data, to reduce the storage space required and to reduce the noise that occurs, it can be provided that the processing unit replaces the matrix data structure after its creation by a reduced matrix data structure, for rectangular, in particular the entire Matrix data structure covering and non-overlapping image areas of the same size in the matrix data structure in each case determines a mean distance value and assigns this mean distance value corresponding to its location pixel of the reduced matrix data structure.

• a) in der im Höhenprofil, insbesondere in der Matrixdatenstruktur oder in der Punktwolke, eine Anzahl von möglichen Betrachtungsbereichen, vorab vorgibt,
• b) für das jeweilige Segment die jeweilige Atemtiefe unter Zugrundelegung jedes der möglichen Betrachtungsbereiche ermittelt, und
• c) denjenigen vorgegebenen möglichen Betrachtungsbereich als Betrachtungsbereich auswählt und heranzieht, für den die ermittelte Atemtiefe am größten ist.

An advantageous method for determining the viewing area provides that the processing unit determines the viewing area by

• a) predetermines in the height profile, in particular in the matrix data structure or in the point cloud, a number of possible viewing areas,
• b) determines the respective breathing depth for each segment on the basis of each of the possible viewing areas, and
• c) selects and uses the given possible viewing area as the viewing area for which the determined breathing depth is greatest.

• a) im Höhenprofil, insbesondere in der Matrixdatenstruktur oder in der Punktwolke, mittels Objekterkennung nach Bereichen sucht oder Mittel zur Auswahl von Bereichen zur Verfügung stellt, die einem menschlichen Kopf und Torso entsprechen, und
• b) den den Torso abbildenden Bereich der des Höhenprofils, insbesondere der Matrixdatenstruktur oder der Punktwolke, oder ein Teilbereich dieses Bereichs als Betrachtungsbereich auswählt, wobei sie insbesondere
• – zur Detektion des Verlaufs oder der Atemtiefe der Brustatmung der dem Kopf nahe Teilbereich dieses Bereichs als Betrachtungsbereich auswählt und/oder
• – zur Detektion des Verlaufs oder der Atemtiefe der Bauchatmung der dem Kopf ferne Teilbereich dieses Bereichs als Betrachtungsbereich auswählt.

A second advantageous method for determining the viewing area, which takes particular account of anatomical conditions, provides that the processing unit determines the viewing area by

• a) in the height profile, in particular in the matrix data structure or in the point cloud, searches for areas by means of object recognition or provides means for selecting areas corresponding to a human head and torso, and
• b) selecting the torso-imaging region of the height profile, in particular the matrix data structure or the point cloud, or a subregion of this region as the viewing region, in particular
• - For detecting the course or the breathing depth of the chest breathing of the near the head portion of this area selects as a viewing area and / or
• - For detecting the course or the breathing depth of the abdominal breathing of the head remote portion of this range selects as a viewing area.

A rapid adaptation of the viewing area can be achieved by the fact that the processing unit for each segment to be examined or for individual height profiles of a segment of the Viewing range by object recognition to the present in the segment height profiles based on the position of the selected viewing area in the respective previous segment adaptively redetermined.

• a) für sämtliche Punkte des Höhenprofils, insbesondere für alle Einträge der Matrixdatenstruktur, separat die Varianz der Werte über das jeweilige Segment bestimmt, und
• b) einen Bereich oder mehrere Bereiche mit jeweils zusammenhängenden Punkten des Höhenprofils, deren jeweilige Varianz oberhalb eines unteren Schwellenwerts oder innerhalb eines vorgegebenen Intervalls liegen, als Betrachtungsbereich auswählt, insbesondere
• – indem sie zusammenhängende Bereiche als Betrachtungsbereich auswählt, deren Größe eine vorgegebene Anzahl von Punkten übersteigt
• – wobei insbesondere ein Bereich dann als zusammenhängend gilt, wenn jeder Punkt des Höhenprofils ausgehend von einem Punkt innerhalb dieses Bereichs über jeweils benachbarte Pixel erreichbar ist,
• – wobei vorzugsweise Punkte des Höhenprofils dann als benachbart gilt,
• – wenn sie durch benachbarte Einträge der Matrix-Datenstruktur festgelegt sind oder
• – wenn sie oder ihre Projektion auf eine horizontale Ebene im Raum einen einen Schwellenwert unterschreitenden Abstand voneinander aufweisen.

A third advantageous method for determining the viewing area, which is numerically stable and efficiently executable, provides that the processing unit determines the viewing area by

• a) for all points of the height profile, in particular for all entries of the matrix data structure, separately determines the variance of the values over the respective segment, and
• b) selects a region or a plurality of regions, each having contiguous points of the height profile whose respective variance is above a lower threshold value or within a predetermined interval, as the viewing region, in particular
• - by selecting contiguous areas as a viewing area whose size exceeds a given number of points
• In which, in particular, an area is deemed to be connected if each point of the height profile can be reached from a point within this area via respectively adjacent pixels,
• - preferably points of the height profile is then considered adjacent,
• If they are defined by adjacent entries of the matrix data structure or
• - If they or their projection on a horizontal plane in space have a threshold below the threshold distance from each other.

In order to be able to detect situations in which the person to be examined moves and to reject the erroneous measured values recorded in these situations, it can be provided that the processing unit separately determines the viewing area in two or more consecutive segments and discards a height profile and, if necessary issues an error message if the viewing area, in particular its size and / or its center of gravity, shifts by a predetermined threshold relative to the respective preceding segment.

• a) das Audio-Signal in eine Anzahl von Audio-Segmente, insbesondere mit einer Länge von 100 ms bis 1 Sekunde, unterteilt und für jedes der Audio-Segmente untersucht, ob in diesem menschliche Atemgeräusche oder andere Geräusche zu hören sind, und für jedes Audio-Segment jeweils ein Klassifikationsergebnis zur Verfügung hält,
• c) zeitgleich aufgenommene Audio-Segmente und Segmente einander zuordnet, und
• d) nach Audio-Segmenten oder Segmenten mit geringer Atemtiefe oder mit fehlenden Atemgeräuschen sucht und die einem solchen Audio-Segment oder Segment zugeordneten Segmente oder Audio-Segmente ebenfalls auf das Vorliegen geringer Atemtiefe oder fehlender Atemgeräuschen untersucht, und
• e) für den Fall, dass in einander zugeordneten Segmenten und Audio-Segmenten jeweils geringe Atemtiefe und fehlender Atemgeräusche detektiert wurden, das Fehlen der Atmung der Person feststellt.

In addition, an acoustic control may be provided for improved detection of respiration. Here, it is provided that the processing unit is preceded by a microphone which at the same time holds the sound emitted by the person in the form of an audio signal at its output parallel to the distance of the person and this audio signal is fed to the processing unit. and that the processing unit

• a) the audio signal divided into a number of audio segments, in particular with a length of 100 ms to 1 second, and examined for each of the audio segments, whether in this human breath sounds or other sounds are heard, and for each Audio segment each has a classification result available,
• c) assigns simultaneously recorded audio segments and segments, and
• d) searches for audio segments or segments with low breathing depth or with no breath sounds and also examines the segments or audio segments assigned to such an audio segment or segment for the presence of low breathing depth or lack of breathing noise, and
• e) in the event that in each associated segments and audio segments each low breathing depth and lack of breath sounds were detected, the lack of respiration of the person detects.

• a) ein Klassifikationssignal erstellt, das für jedes Audio-Segment jeweils zumindest ein Klassifikationsergebnis enthält, das die Art und Stärke des jeweiligen Geräuschs im Zeitbereich des jeweiligen Audio-Segment angibt,
• b) jeweils eine Anzahl von aufeinander folgenden Audio-Segmente zu weiteren Audio-Segmenten, insbesondere mit einer Dauer von 5 bis 10 Sekunden zusammengefasst, die vorzugsweise dieselben Zeitbereiche umfassen wie die Segmente,
• c) über die innerhalb eines weiteren Audio-Segments enthaltenen Klassifikationsergebnisse eine Mittelung vornimmt und dem weiteren Audio-Segment den jeweilige Mittelwert zuweist,
• d) ein weiteres Klassifikationssignal durch Interpolation der Mittelwerte der weiteren Audio-Segmente erstellt,
• e) im weiteren Klassifikationssignal sowie im Atemtiefesignal nach Zeitpunkten sucht, zu denen in den beiden Signalen starke Änderungen auftreten, und die so erkannten Zeitpunkte als umso relevanter eingestuft, je stärker die Änderung des jeweiligen Signals zum jeweiligen Zeitpunkt ist, wobei sie diesen Zeitpunkten jeweils ein diesbezügliches Relevanzmaß zuordnet, und
• f) jene Zeitpunkten als Startpunkte oder Endpunkte eines Atemstillstandes detektiert, für welche der Betrag des Relevanzwerts über dem Betrag eines Schwellwert liegt, wobei sie insbesondere
• – der Schwellenwert bildet, indem sie über einen Zeitbereich, insbesondere vor und/oder nach dem Vergleichszeitpunkt mit dem Referenzmaß ein Mittelwert des Referenzmaßes bildet und der Schwellenwert im Bereich zwischen 120% und 200% dieses Mittelwerts festlegt.

A particularly preferred embodiment of the invention, which allows detection of the beginning and end of respiratory misfires, provides that the processing unit

• a) generates a classification signal which contains for each audio segment at least one classification result which indicates the type and strength of the respective noise in the time domain of the respective audio segment,
• b) in each case a number of successive audio segments are combined to form further audio segments, in particular having a duration of 5 to 10 seconds, which preferably comprise the same time ranges as the segments,
• c) averaging over the classification results contained within a further audio segment and assigning the respective average value to the further audio segment,
• d) creating a further classification signal by interpolation of the mean values of the further audio segments,
• e) in the further classification signal as well as in the breath depth signal searches for times at which strong changes occur in the two signals, and the times thus identified as all the more relevant, the stronger the change of the respective signal at the respective time, in each case these times assigns relevance measure to this, and
• f) detects those points in time as starting points or end points of a respiratory arrest for which the magnitude of the relevance value is above the magnitude of a threshold, in particular
• - The threshold forms by over a time range, in particular before and / or after the comparison time with the reference measurement forms an average of the reference measure and the threshold in the range between 120% and 200% of this average.

Several embodiments and variants of the invention will be described in more detail with reference to the following drawing figures. 1 shows a bed with a sleeping person. 2 shows a view of the in 1 shown bed from above. 2a shows a waveform of the signal in a person with normal breathing. 2 B shows a waveform of the signal in a person with a respiratory obstruction. In 3 is a sequence of signals of temporally successive segments shown. 4 and 5 show a first procedure for determining the viewing area. 6 shows an alternative procedure for determining the viewing area with the aid of object recognition algorithms. 7 and 8th show a third procedure for determining the viewing range. 9 shows a combination of a method based on the evaluation of height profiles with an acoustic method for the detection of respiratory misfires.

In 1 is a bed 10 shown from the side. The bed 10 has a body, a mattress 11 , a pillow 13 and a blanket 12 , In bed 10 is a sleeping person 1 whose head is on the pillow 13 and his torso and legs are off the comforter 12 are covered.

Above the person is a detector unit 20 arranged with the distance of the person 1 or from a plurality of points on the person of a given, relative to the detector unit 20 can be determined. The detector unit 20 is a processing unit 50 downstream, which performs the numerical processing steps shown below.

In the present case, the detector unit is 20 to a unit, each the normal distance of a point on the person 1 from one above the person 1 horizontally extending reference plane 21 indicates. The detector unit 20 measures the respective distance d ₁ ,..., d _n at n different positions, which in the present exemplary embodiment are arranged in the form of a grid in the form of rows and columns, and uses these distance measurement values to produce a height profile H.

In the present embodiment, to create a height profile H, a detector unit 20 used, which is located about 2 meters above the person. Such a detector unit can be constructed in different ways.

In a first embodiment, the detector unit can be designed as a time-of-flight camera (TOF camera). This determines the distance to an object is determined by the "flight time" of a transmitted light pulse. This results in distance measurements with a lateral resolution, which typically moves approximately in the range of 320 × 240 pixels. The concrete operation of such a TOF camera is known from the prior art and is in Hansard, M., Lee, S., Choi, O., and Horaud, R. (2013) , Time-of-fight cameras. Springer shown in more detail.

Alternatively, a height profile can also be determined by means of light-section method. These are made using a light source that focuses on the person 1 directed the surface is triangulated. You do not get absolute distance measurements, but only a height profile. Since, in the context of the invention, no absolute measured values are used in any case, but only the relative values such as variance or the change of the variance over time, height profiles can also easily be used for the invention. The following publication describes a system that provides height profiles at a frame rate of approximately 40 fps and at a resolution of 640 × 480 pixels: Oike, Y., Shintaku, H., Takayama, S., Ikeda, M., and Asada, K. (2003) , Real-time and high resolution 3d imaging system using light-section method and smart cmos sensor. In Sensors, 2003. Proceedings of IEEE, volume 1, pages 502-507. IEEE ,

Another alternative method for taking a height profile uses a radar measurement. For this purpose, radar measuring devices, possibly in their direction controllable Radarmesseinrichtungen, so-called phased arrays are used. By means of phase shifts in the radar pulse in the antenna array, the impulse can be at some point on the body of the person 1 be steered and thus the room are scanned. In principle, a high spatial resolution can be achieved with such a radar measuring device. It is also easily possible to have 30 height profiles per second achieve. The concrete structure of such Radarmesseinrichtungen is in Mailloux, RJ (2005) , Phased array antenna handbook. Artech House Boston.

2 shows a view of the in 1 shown bed 10 from above, wherein at a number n of positions in each case the distance d ₁ , ..., d _{n of} the person 1 and / or the person 1 covering duvet 12 and / or the pillow 13 from the reference plane 21 is measured. The individual positions at which measurements are taken are arranged in a grid pattern and in 2 represented by crosses. These positions can be clearly described by specifying an x-coordinate and a y-coordinate. From the individual determined distances d ₁ ,..., D _n , a height profile H of the person is continuously recorded at successive recording times t ₁ ,..., T _p 1 created. In the present case, the height profile contains the individual determined distances d ₁ ,..., D _n and is kept available in the form of a matrix data structure.

However, such a representation of the height profile is not mandatory for the invention for several reasons.

The height profile need not necessarily be at points the distance of the surface of the person or the surface of an object located on or next to the person to a reference plane 21 give, which are arranged laterally grid-shaped. Rather, it is also conceivable that the distances along predetermined radiation, which are measured by the detector unit to be measured and then the vertical distances from the respective intersection with the surface to the reference plane using the measured distance and the respective used measuring angle are calculated. The angles of the various measuring beams can be chosen so that when hitting a plane parallel to the reference plane 21 is a grid-shaped arrangement would result. Even with this very special selection of measuring beams, lateral measuring points, whose x and y coordinates are derived from the regular grid arrangement in FIG 2 differ as soon as the measured surface deviates from a plane. However, since the variation of the height of the measured surface of an object located on or near the person in relation to the distance of the detector unit is typically small, the lateral coordinates in the above-mentioned arrangement of the measuring beams can be approximately in the arrangement of 2 and thus assumed in a matrix data structure.

In general, the individual distance values respectively give the distance of the point of intersection of a beam which has been predetermined in advance relative to the detector unit, and in particular emanates from the detector unit, with the surface of the person ( 1 ) or the surface of one on or next to the person ( 1 ) from a reference point or reference plane ( 21 ) at.

Furthermore, it is not necessary for the individual available distance values to be arranged in a matrix data structure. It is also possible that a data structure with a different structure is selected and only distances along specific rays are determined.

In the context of the invention, it is also possible to specify the height profile in the form of a point cloud. The point cloud comprises a list of points each indicating their respective coordinates in three dimensions relative to a reference point in space.

Also other representations of the relative position of the surface of the person 1 or the surface of one on or next to the person 1 The object in relation to the detector unit or to the room is possible if individual points on this surface can be determined.

The one from the detector unit 20 the area covered or covered by the positions 25 is limited to the body of the person 1 To minimize the number of irrelevant movements of the person 1 or to detect motion artifacts.

The distance measurement values d ₁ ,..., D _{n of} the height profile H will be referred to below as H (x, y, t), the first two coordinates denoting the spatial position and the last coordinate t the recording time t ₁ ,... , t _p . It is assumed that x ε [0, ..., X] and y ε [0, ..., Y], that is, the spatial resolution of the data stream is X × Y. The third coordinate t stands for the time and refers to the recording time t ₁ , ... t _p , in multiples of the temporal sampling density or sampling rate, z. B. 1/30 s for 30 frames / second.

Overall, a three-dimensional data structure is in the course of recording created in each recording time t _1, ..., t, _p is each a two-dimensional matrix data structure available, the Entries each the distance of the person 1 from the reference plane 21 in an area defined by the position of the entries in the matrix data structure. Each matrix data structure created in this way is the same size. All matrix data structures each contain memory positions for the individual distance measurement values d ₁ ,..., D _n or for values derived therefrom.

In an optional step, the spatial resolution of the matrix data structure can be reduced by combining a plurality of entries of the matrix data structure into an entry of a reduced matrix data structure. In the simplest case, this can mean that only individual ones of the distance measurement values d are used for the formation of the reduced matrix data structure and the other distance measurement values are discarded. For integer parameters a and b, one obtains a reduced matrix data structure H _r (x, y, t) = H (ax, by, t) whose memory requirement is smaller than the memory requirement of the matrix data structure by the (a × b) -fold.

To get a better and more robust result, before reducing the resolution, the matrix data structure can be smoothed or filtered. For this, one calculates the two-dimensional Fourier transform of H with respect to the first two coordinates multiplied by a filter transfer function, periodicizes this signal with the parameters X / a and Y / b and then calculates the inverse Fourier transform to obtain the reduced matrix data structure. The subsampling described above is merely a special case thereof with a constant filter transfer function. An averaging over a number of within rectangular areas of the matrix data structure of size a × b can also be realized by means of a filtering followed by subsampling.

In a further step, the recording interval in which height profiles H were created is subdivided into preferably contiguous and non-overlapping time intervals of 5 to 10 seconds. Segments S ₁ ,..., S _{m are} created which contain the individual determined matrix data structures that were recorded within a time interval. Alternatively, it is also possible that the individual time segments overlap or that height profiles of individual time domains are not contained in any of the segments S ₁ ,..., S _n .

For each of the segments S _i , a data structure H _i (x, y, t) results, where ε [0,..., X], y ∈ [0,..., Y] and t ∈ S _i . In addition, the times at which a segment begins are designated T _{3D, i} . In the following it is described how for each of these time blocks respiratory activity can be determined and optionally also the depth of breath can be measured.

For each segment S _i becomes a viewing area 22 selected, which indicates the region or regions of the matrix data structure in which a breathing movement is expected. This can be preset by hand and can be selected constantly over the individual segments S _i .

However, there is also the possibility of the viewing area 22 to adapt to movements of the person. However, since the individual segments S _i only contain matrix data structures from time periods of about 10 seconds, an adaptation within a segment S _{i is} not required in most cases. Within a segment S _i , the viewing area becomes 22 usually assumed constant.

The adaptation of the viewing area 22 For the individual segments S _i has the advantage that easy detection of the respiration of the person 1 is also possible if this moves in sleep. The viewing area becomes advantageous 22 automatically determined. Below, three different techniques are described to the viewing area 22 for each segment S _i automatically adapt to possible movements of the person 1 to be able to follow.

After the viewing area 22 is fixed, a temporal signal s _{i is} created, in which for each matrix data structure in each case a signal value s _i (t) is determined. The signal value is determined in the present case by determining the mean value of the distance values H (x, y, t) for the respective time t, their position within the viewing range 22 lies. This gives a signal s _i (t) which depends only more on time.

This signal s (t) may contain noise. In order to obtain from this noisy signal s (t) the respiration, and optionally also a measure of the respiratory strength, this signal s (t) can be subjected to noise filtering. An advantageous noise suppression can be achieved, for example, by suppressing signal components with a frequency of more than 1 Hz. Alternatively or additionally, it may also be provided to suppress DC components or frequencies up to 0.1 Hz. With this filtering can be achieved that only frequencies that are relevant for the determination of respiration are left. The signal s _i (t) thus obtained agrees well with the actually performed breathing movements.

The breathing depth T can be derived particularly advantageously from this signal s (t). There are many possibilities for this, two of which are described in more detail. 2a shows a first waveform of the signal s (t), wherein within the waveform for a minimum s _min and a maximum s _{max is} searched. Subsequently, the difference Δ between the minimum s _min and a maximum s _{max is} formed and this difference Δ is used as a measure for the breathing depth T.

How out 2 B Obviously, is the breathing of the person 1 much weaker than the one in 2a represented signal. For this reason, the difference Δ between the minimum s _min and a maximum s _{max is} smaller than in the 2a represented signal.

If a plurality of local maxima and minima are found within a signal, it is also possible to use the greatest difference between a minimum and the respectively following local maximum as the breathing depth T.

Alternatively, it is also possible for the signal to be subjected to a spectral transformation, in particular Fourier transformation, cosine transformation or wavelet transformation. Within a given frequency band, in particular from 0.1 Hz to 1 Hz, the spectral component with the highest signal energy is searched for. The signal energy of this spectral component is used to characterize the breathing depth T in this segment.

In 3 is a sequence of signals s _i (t) of temporally successive segments S _i shown, wherein during the segments S ₃ and S ₄ a breathing misfire the person 1 is present. Furthermore is in 3 the respiratory depth A _i shown in each case also on the same time axis for each segment S _i as well as an interpolation curve T (t) indicating the depth of breathing T for all time points within the illustrated segments S _1, ..., S. ₇ It can clearly be seen that the values A ₁ ,..., A ₇ for the respiratory depth during the respiratory misfire in segments S ₃ and S ₄ are markedly reduced compared with the other segments.

The following are individual possible ways to automatically detect the viewing area 22 shown within a segment S _i .

In a first approach ( 4 ) individual possible viewing ranges R _1,1 , ... are given. Subsequently, in each case the breathing depth T is determined in all possible viewing ranges R _1,1 ,... According to the methods mentioned above. In 5 For example, individual signals s _3,4 , s _2,3 and s _{3,3 are} shown, which were created on the basis of the possible viewing ranges R _3,4 , R _2,3 and R _3,3 . Since the possible viewing area R _3,3 exactly the chest area of the person 1 The signal amplitudes in this possible viewing area R _{3,3 are also} greater than in the other two viewing windows R _3,4 , R _2,3 , which are the person's head and arms, respectively 1 depict. For this reason, the possible observation range R _3,3 becomes the observation range 22 selected for the respective segment S _i .

The position and size of the viewing area R _3,3 can be further improved by attempting to shift the vertices of the rectangle defining the viewing area and recomputing the breathing depth T based on this modified viewing area. By varying the corner points of the viewing area, this can be adaptively improved until it is no longer possible to increase the determined breathing depth by displacing the corner points.

Another, in the 6 illustrated procedure for determining the viewing area 22 can be done by means of object recognition. First, in one or more of the matrix data structures of the segment S _i, by means of object recognition, regions are determined 31 . 32 searched, which correspond to a human head and torso or torso. The torso imaging area 32 The matrix data structure or a subarea of this area can be used as a viewing area 22 to be selected.

For the detection of the chest breathing the head can 31 near subarea 33 this area 32 as a viewing area 22 to be selected. For the detection of the abdominal breathing can be the head 31 distant subarea 34 this area 32 as a viewing area 22 to be selected.

For the object recognition several different image processing methods come into question. The object of the known object recognition methods is to automatically or outlines the contours of a human body Semi-automatic, ie partly with human support, to recognize in a 3D image or height profile. Such a procedure is described in the following publications:
Gabriele Fanelli, Juergen Gall, and Luc Van Gool. Real time head pose estimation with random regression forests. In Computer Vision and Pattern Recognition (CVPR), pages 617-624, June 2011 ,
Jamie Shotton, Ross Girshick, Andrew Fitzgibbon, Toby Sharp, Mat Cook, Mark Finocchio, Richard Moore, Pushmeet Kohli, Antonio Criminisi, Alex Kipman, et al. Efficient human pose estimation from single depth images. Pattern Analysis and Machine Intelligence, IEEE Transactions on, 35 (12): 2821-2840, 2013 ,

In the detection of body structures by means of object recognition statistical learning methods are used. While the first publication aims to determine the posture of the head, the second publication aims to localize the body and, in particular, the joints. With both methods, the position of the head and the torso in the height profile H can be determined.

Another possibility is to first manually set the height profile H of the head and then to determine in each segment S _i by means of a correlation where the head is at that time. In the same way, the torso can also be found in each segment S _i or in each matrix data structure within a segment S _i . If the position of the torso and head is known, it is easy to divide the torso into the chest and abdominal region. This allows one to separately detect respiration both in the chest and in the abdomen. The same segmentation of the person's body can be realized using many different algorithms.

Can in one of the segments S _i a viewing area 22 can be identified, in most cases with little effort of the viewing area 22 in the segment S _{i + 1} which immediately follows this segment S _i . In both of the above-described procedures, knowledge of the approximate position of the image of the abdominal or thoracic region facilitates locating the abdominal or thoracic region in the respective next segment S _{i + 1} .

In 7 and 8th is a third way to find the abdominal or chest area of the person as a viewing area 22 shown. In a first step, the variance of the values over the respective segment S _{i is} determined separately for all entries of the matrix data structure. Subsequently, the region or contiguous regions of entries within the matrix data structure becomes a viewing region 22 whose respective variances are above a lower threshold or within a predetermined interval. 7 shows a section through a height profile of a person along the section line VII-VII 2 , The dashed line represents the time average of the individual distance values at the respective position. The arrows show the variance at the respective position. It turns out that in the area of the chest and in the area of the person's belly more variances occur than in the rest of the body. These areas are called viewing areas 22a . 22b selected.

In order to be able to avoid misdetection by individual artefacts, only contiguous areas are used as the viewing area 22a . 22b of the matrix data structure whose size exceeds a predetermined number of entries.

This variant of determining the viewing area 22 also allows the detection of position changes of the person as well as measurement errors. Will the viewing area 22 in two or more consecutive segments S _i , ... determined separately, so a height profile can be discarded and an error message is issued when the size and the center of gravity of a viewing area 22a . 22b , opposite the corresponding viewing area 22a . 22b in the respective preceding segment S _i shifts or changes by a predetermined threshold.

Of course, it is also possible to use the aforementioned procedures for determining the viewing range 22 to combine. For example, using the image-recognition methods for object recognition, the torso can be determined and then within that range only those entries can be selected for which a significant movement was detected using the variance. Alternatively, it is also possible to have several different possible viewing ranges 22 and then that viewing area 22 , which was used to determine the greatest breathing depth, should be used for respiratory detection.

In the illustrated embodiments, a height profile was always used, which is stored in a matrix data structure. However, the invention also allows alternative embodiments in which the height profiles each set only a number of points in space. It is not necessary that the respective points be explicitly determined by coordinates as long as the respective point is due the information in height profile H in space is clearly defined. In the present embodiments, this is implied by the concrete position at which the distance of the person 1 is determined, given.

Alternatively, it is also possible to specify the height profile by means of a point cloud, ie to essentially specify a list of points in the space. In such a case, the viewing area 22 be defined as a three-dimensional area located in the chest or abdomen of the person 1 located. For the formation of the respective mean only points are used, which are within the respective viewing range 22 are located. To all elevations and subsidence of the chest or abdomen of the person 1 can cover, as a viewing area 22 an area may be selected that also includes a volume above the chest or abdomen area to about 10 cm above the chest or abdominal area.

A preferred embodiment of the invention also offers the possibility of breathing depth and breathing of the person 1 parallel to the height profiles by acoustic recordings. In 1 is an optional microphone 40 represented, with which an audio signal s _{a is} recorded. The microphone 40 created audio signal s _a is digitized and stored at a predetermined sampling rate and kept available. The microphone is connected to the processing unit 50 connected.

The audio signal s _a is first divided into audio segments S _a with a predetermined length of about 200 ms. These audio segments S _a may generally also overlap, but in the present case contiguous, non-overlapping audio segments S _{a are} selected.

Optionally, the audio segments S _a may be multiplied by a window function to avoid edge effects in the Fourier methods described below. In addition, a time T _{audio, i is also} associated with each audio segment S _a , which indicates at which time the segment f _i starts.

Each of the audio segments S _Ai is supplied to a classification with which it is detected whether in the respective audio segment S _Ai no noise, a background noise H, a respiratory noise or a snore sound A / S can be heard. For each audio segment S _Ai , a classification result is produced which, depending on whether no noise, a background noise, a respiratory noise or a snoring sound can be heard, has a corresponding value N, H, A / S ( 9 ) having. The classification results are combined into a discrete temporal classification signal A (t).

The following describes how to use statistical learning methods to examine each of these individual audio segments S _Ai for the presence of snoring sounds.

In a first step, a feature vector m _i can be extracted from the audio segments S _Ai to detect snoring sounds in the individual audio segments S _Ai . The feature vector m _i for the ith segment audio segment S _Ai is calculated directly from the samples of the respective audio segment S _Ai . The i-th feature vector m _i can be of various dimensions, depending on which methods are used to calculate the features. Below are a few different techniques that can be used to generate features.

By means of the Fourier transformation of an audio segment S _Ai , a spectral analysis of the respective audio segment S _Ai can be carried out. As characteristics, the energies in certain frequency bands are used, that is the sum of the squares of the Fourier coefficients over certain prespecified bands. This gives, for each audio segment S _Ai, a vector with a length determined by the number of bands. If this vector is discretely cosine transformed once again, one obtains another possible set of coefficients of the feature vector m _i .

Another possible feature of a feature vector m _i is the energy in an audio segment S _Ai . In addition, the number of zero crossings, that is to say the number of sign changes in an audio segment S _{Ai, can also be used} as a possible feature of a feature vector m _i .

After determining how the feature vectors m _{i are} composed, a statistical learning method is selected. There are a large number of possible methods available in the state of the art for this, therefore, the various options will not be discussed in more detail here. These methods make it possible to identify an audio segment S _Ai as a snore noise on the basis of a feature vector m _i . For this purpose, the feature space is subdivided into a set of points that are assigned to the snore and the remainder, which is classified as non-snoring. In order to be able to perform this classification, these algorithms are first trained on the basis of known snore sounds, that is, they are manual Audio segments are selected with snoring sounds, the feature vectors m _i calculated by it and then the classification algorithm is trained with it. Implementations of all statistical learning methods used are freely available and already implemented.

Using the above-mentioned classification methods, it is possible to determine which audio segments S _Ai contain snoring sounds. In an average snore signal, however, quiet periods are often between the loud phases, as often snored only by inhalation, as in 9 is shown. To correctly snore the entire snoring phase as well as resting segments between the loud episodes, several segments before and after are considered at the i-th point. If there are more segments with snoring noise in this quantity than a certain threshold, snoring is detected at time T _{audio, i} .

The goal of classifying recorded sounds to distinguish breath sounds such as snoring and breathing from other background sounds that may occur, can be solved by methods known in the art. Such methods are known in particular from the following publications:
Hisham Alshaer, Aditya Pandya, T Douglas Bradley, and Frank Rudzicz. Subject independent identification of breath sounds components using multiple classifiers. In Acoustics, Speech and Signal Processing (ICASSP), 2014 IEEE International Conference on, pages 3577-3581. IEEE, 2014 ,

Various statistical learning methods are used to divide the recorded sounds into different categories. In the first paper, the microphone is fixed in a mask on the face of the patient, so there are hardly any background noise, or these are negligible. In this publication, the recorded sounds are inhaled, exhaled, snore and wheeze. It is generated to a variety of different characteristics, with which the selected learning method works.
M Cavusoglu, M Kamasak, O Erogul, T Ciloglu, Y Serinagaoglu, and T Akcam. An efficient method for snoring / non-classification of sleep sounds. Physiological measurement, 28 (8): 841, 2007 ,

Also in this publication, statistical learning methods are used to divide the recorded sounds into different categories. The microphone is about 15 cm from the subject's head and the recorded sounds are divided into "snoring" and "not snoring".

In 9 are snore sounds X _s shown exhaling, between these snoring sounds X _s and background noise X _{H are} recorded. On the basis of the illustrated classification, each of the audio segments S _Ai can be assigned a classification value K _a , which in the uppermost region of the 9 is shown.

Optionally, depending on the quality of the audio signal and levels of background noise, the snoring sounds can be more accurately subclassed, such as sub-classes. B. pathological and non-pathological. Under very favorable conditions, breathing can also be classified as a separate noise and its strength can be determined.

In 9 is the result of an improved embodiment of the invention shown in more detail, in which both the results of the evaluation of the height profiles as well as the results of the acoustic snoring analysis enter into the evaluation of the breathability. Typically, the segments S _i and the audio segments S _Ai are not time correlated with each other.

For this reason, simultaneously recorded audio segments S _Ai and segments S _i are initially assigned to one another on the basis of their respective recording time. For each segment S _{i, there} are typically several audio segments S _Ai available that were recorded during the period of the segment.

In the present embodiment of the invention, segments S _i with low breathing depth T are searched for. The in 9 Data shown can be found that in the segment S ₃ a breathing depth T is present, which is substantially less than in the two preceding segments S ₁ , S ₂ . Subsequently, in all this segment S ₃ associated audio segments S _Ai searched for lack of breath sounds. Here it can be stated that in the audio segments S _Ai , which were recorded simultaneously with the segment S ₃ , no snoring or breathing noises but only background noise X _H available. Since, on the one hand, a low level of breathing was recognized and, on the other hand, no snoring or breathing sounds could be detected, it can be assumed that the person in question is not breathing.

From the data extracted from the audio signals and height profiles, statements about respiration can now be made at any time. The procedure is as follows: Two indices I and J are sought, so that the respective start times T _{audio, I of} the audio segments S _Ai and the start times T _{3D, J of} the segments S are closest to the time t at all possible times that were selected for audio segmentation and segmentation of the elevation profiles. The possibilities that arise at this time are summarized in the following table: Snoring at T _{audio, I} No snoring at T _{audio, I} great breathing depth at T _{3D, J} breathing breathing no or little breathing depth at T _{3D, J} breathing Apnea

In the following, a further embodiment of the invention is shown in which an improved combination of the determined breathing depth T and the acoustic classification of the breathing sounds and thus an improved detection of respiratory misfires is made possible. By interpolation of the determined breathing depths over time, a respiratory depression signal T (t) can be determined.

This classification is based on the respiratory depression function T (t), which indicates the time course of the respiratory depth, as well as the classification signal A (t), which indicates at each time whether and, if so, which respiratory sounds were present. In the present exemplary embodiment, the classification signal A (t) is modified and defined as a vector-valued function which assigns to each time point a vector whose components in each case indicate the intensity of the respective detected signal. Concretely, the classification signal A (t) could comprise components, one of which represents the strength of the snoring sound or a snore sound of a certain type, while another represents the strength of the background noise present at the time.

It is then possible to combine several audio segments to form another audio segment with a length of, for example, 5 to 10 seconds and to determine the respective signal strengths in this further audio segment by means of an averaging. With this step, it is possible to classify a total of snoring for longer periods in which a person snores while inhaling but exhales quietly. This averaging produces an averaged classification signal B (t). It is particularly advantageous if the averaged classification signal B (t) and the respiratory depression signal A (t) are defined on segments which coincide with one another or on further audio segments.

In a subsequent step, both the respiratory depression signal T (t) and the further classification signal B (t) are searched for times at which strong changes occur in the two signals T (t), B (t). These are, for example, times at which breath sounds disappear or re-insert or respiratory movements disappear or re-insert.

The times thus recognized are considered to be all the more relevant, the stronger the change in the signal T (t), B (t) at the respective time. A REL relevance value is created that indicates the relevance of each point in time for the beginning or end of a low-respiration phase.

Those points are selected as starting points R _S or end points R _{E of} a respiratory arrest, for which the amount of the relevance measure REL is above the magnitude of a predetermined threshold value. A start time R _S can be assumed when the relevance value REL is negative, an end time R _E can be assumed when the relevance value REL is positive. The period AP between the start time R _S and the end time R _E is regarded as a period of low or absent breathing activity.

In particular, the threshold value can be formed in a sliding manner by establishing a mean value of the reference measurement over a time range, in particular before and / or after the comparison time with the reference measurement, and setting the threshold in the range between 120% and 200% of this mean value.

Overall, it is within the scope of the invention possible to record the entire sleep of a person over several hours and the height profiles, possibly also the audio signals, into individual segments, and possibly also audio segments. Of course, it is also possible to consider only individual particularly relevant sleep phases of the person and to assign only a few of the recorded height profiles or audio samples to individual segments and possibly audio segments.

Furthermore, it is possible to record the number and duration of the determined phases of low breathing depth or lack of respiratory activity over the entire sleep and to submit statistical analyzes.

QUOTES INCLUDE IN THE DESCRIPTION

This list of the documents listed by the applicant has been generated automatically and is included solely for the better information of the reader. The list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions.

Cited non-patent literature

• Hansard, M., Lee, S., Choi, O., and Horaud, R. (2013) [0050]
• Oike, Y., Shintaku, H., Takayama, S., Ikeda, M., and Asada, K. (2003) [0051]
• Sensors, 2003. Proceedings of IEEE, volume 1, pages 502-507. IEEE [0051]
• Mailloux, RJ (2005) [0052]
• Gabriele Fanelli, Juergen Gall, and Luc Van Gool. Real time head pose estimation with random regression forests. In Computer Vision and Pattern Recognition (CVPR), pages 617-624, June 2011 [0082]
• Jamie Shotton, Ross Girshick, Andrew Fitzgibbon, Toby Sharp, Mat Cook, Mark Finocchio, Richard Moore, Pushmeet Kohli, Antonio Criminisi, Alex Kipman, et al. Efficient human pose estimation from single depth images. Pattern Analysis and Machine Intelligence, IEEE Transactions on, 35 (12): 2821-2840, 2013 [0082]
• Hisham Alshaer, Aditya Pandya, T Douglas Bradley, and Frank Rudzicz. Subject independent identification of breath sounds components using multiple classifiers. In Acoustics, Speech and Signal Processing (ICASSP), 2014 IEEE International Conference on, pages 3577-3581. IEEE, 2014 [0102]
• M Cavusoglu, M Kamasak, O Erogul, T Ciloglu, Y Serinagaoglu, and T Akcam. An efficient method for snoring / non-classification of sleep sounds. Physiological measurement, 28 (8): 841, 2007 [0103]

Claims (32)

Method for determining the time course of the breathing depth (T) of a person, in particular a sleeping person, with a person ( 1 ) directed detector unit ( 20 ) continuously to successive recording times (t ₁ , ..., t _p ) each have a height profile (H) of the person ( 1 a) wherein in the height profile (H) a number of at least two points in the space is defined, which on the surface of the person ( 1 ) or on the surface of one on or next to the person ( 1 b) wherein for each of the recording times (t ₁ , ..., t _p ) the respective height profile (H) is stored in a data structure and kept available, c) where a number of successive recording times ( t ₁ , ..., t _p ), in particular within a time range of 3 to 20 seconds, recorded height profiles (H) to a segment (S ₁ , ..., S _q ) is summarized, d) wherein a range, the the abdominal or chest area of the person as a function of the respective reference point or reference area ( 21 ) indicates as a viewing area ( 22 ) is selected for each height profile (H) within the segment (S ₁ ,..., S _q ) in each case separately the mean value of the distances within the viewing region ( 22 ) points of the height profile (H) to a reference point or reference object is determined, and for the segment (S ₁ , ..., S _q ) a signal (s _i ) is determined to the respective recording time (t ₁ , .. ., t _p ) of the height profile (H) is assigned to the mean value determined for this height profile (H), and f) wherein, based on the determined signal (s _i ), in particular based on its signal amplitude, one or more of the temporal course of the breathing depth (T) characterizing values is or are determined, characterized in that g) the height profile (H) is characterized by a two-dimensional matrix data structure comprising a number of rows and columns, h) a number of positions arranged in rows and columns in grid-shaped positions in which the distance values (d ₁ ,..., d _n ) of the height profile (H) are respectively determined, or - the matrix data structure is created by the distance values (d ₁ ,..., d _n ) a ls the distances along predetermined radiation emanating from the detector unit are measured, and then the vertical distances from the respective intersection with the surface to the reference plane using the measured distance and the respective used measuring angle are calculated, the angles of the various measuring beams so be chosen so that when the measuring beams hit a plane parallel to the reference plane ( 21 i) the matrix data structure has a raster of the same size and structure, j) the matrix data structure is created by adding the distance values (d ₁ , ..., d _n ) recorded at the respective positions to the K) the matrix data structure after its creation is replaced in dimensions by a reduced matrix data structure whose spatial resolution is reduced by adding multiple entries of the matrix data structure to an entry of a reduced matrix data structure in particular, only individual ones of the distance measurement values (d) are used for the formation of the reduced matrix data structure and the remaining distance measurement values are discarded, and wherein preferably the parameters (a, b) are determined as integers and the reduced matrix d ten structure H _r (x, y, t) = H (ax, by, t) is determined. Method for determining the time course of the breathing depth (T) of a person, in particular a sleeping person, with a person ( 1 ) directed detector unit ( 20 ) continuously to successive recording times (t ₁ , ..., t _p ) each have a height profile (H) of the person ( 1 a) wherein the height profile (H) has a number of at least two distance values (d ₁ ,..., d _n ) for defining in each case a point in space, wherein the individual distance values in each case the distance of the intersection point of a relative to Detector unit predetermined, in particular emanating from the detector unit, the beam with the surface of the person ( 1 ) or the surface of one on or next to the person ( 1 ) from a reference point or reference plane ( 21 b) wherein for each of the recording times (t ₁ , ..., t _p ) in each case a data structure is created, which contains the respective height profile (H), wherein all data structures thus created each have the same size and each with memory positions for the c) where a number of height profiles (H) recorded to successive recording times (t ₁ ,..., t _p ), in particular within a time range of 3 to 20 seconds, become a segment (FIG. S ₁ , ..., S _q ) is summarized, d) wherein a number of memory positions of the data structure in which distance values (d ₁ ,..., d _n ) are stored which are the distance of the abdominal or chest region of the person as a function of the respective reference point or reference range ( 21 ), as a viewing area ( 22 ) is selected for each height profile (H) within a segment (S ₁ ,..., S _q ) in each case separately the mean value of the within ( 22 ) distance values (d ₁ , ..., d _n ) is determined, and for the segment (S ₁ , ..., S _q ), a signal (s _i ) is determined, to the respective recording time (t ₁ ,. .., t _p ) of the height profile (H) is assigned to the mean value determined for this height profile (H), and f) wherein based on the detected signal (s _i ), in particular based on its signal amplitude, one or more of the time course of the Resp. Characterizing values are or are characterized, characterized in that g) the height profile (H) is characterized by a two-dimensional matrix data structure comprising a number of rows and columns, h) a number of positions arranged in rows and columns in grid form at which the distance values (d ₁ ,..., d _n ) of the height profile (H) are respectively determined, or - the matrix data structure is created by taking the distance values (d ₁ ,..., d _n ) as the distances along given St are measured and then the vertical distances from the respective intersection with the surface to the reference plane are calculated using the measured distance and the respectively used measuring angle, wherein the angles of the different measuring beams are chosen so that Impact of the measuring beams on a plane parallel to the reference plane ( 21 i) the matrix data structure has a raster of the same size and structure, j) the matrix data structure is created by adding the distance values (d ₁ , ..., d _n ) recorded at the respective positions to the K) the matrix data structure after its creation is replaced in dimensions by a reduced matrix data structure whose spatial resolution is reduced by adding multiple entries of the matrix data structure to an entry of a reduced matrix data structure in particular, only individual ones of the distance measurement values (d) are used for the formation of the reduced matrix data structure and the remaining distance measurement values are discarded, and wherein preferably the parameters (a, b) are determined as integers and the reduced matrix d ten structure H _r (x, y, t) = H (ax, by, t) is determined. A method according to claim 1, characterized in that the height profile (H) describes a point cloud with a number of at least two points in the space, which on the surface of the person ( 1 ) or on the surface of one on or next to the person ( 1 ) are located. Method according to one of the preceding claims, characterized in that for characterizing the breathing depth (T) in a segment (S ₁ , ..., S _q ) from the signal (s _i ) at least one maximum (s _max ) and at least one minimum (s _min ) and the at least one difference (Δ) between the maximum (s _max ) and the minimum (s _min ) as the breathing depth (T) characterizing value for the segment (S ₁ , ..., S _q ) is used. Method according to one of the preceding claims, characterized in that the signal (s _i ) is subjected to a spectral transformation, in particular Fourier transformation or cosine transformation or wavelet transformation, and within a predetermined frequency band, in particular from 0.1 Hz to 1 Hz, the spectral component with the highest signal level is searched for and the signal energy of this spectral component is used to characterize the breathing depth (T) in this segment (S ₁ ,..., S _q ). Method according to one of the preceding claims, characterized in that the signal (s _i ) associated with a segment (S ₁ , ..., S _q ) is subjected to noise filtering after its creation, in particular before the determination of the value characterizing the breathing depth, wherein in particular a) signal components with a frequency of more than 0.5 Hz are suppressed, and / or b) DC components are suppressed. Method according to one of the preceding claims, characterized in that a) that the breathing depth (T) is determined separately for a number of overlapping or non-overlapping segments (S ₁ , ..., S _q ) and / or b) that the viewing area ( 22 ) for each segment (S ₁ , ..., S _q ) separately, in particular on the basis of the respective preceding segment (S ₁ ,..., S _q ) 22 ), is determined. A method according to claim 1, characterized in that the matrix data structure is replaced after its creation in their dimensions by a reduced matrix data structure, wherein for rectangular, in particular the entire matrix data structure covering and non-overlapping image areas of the same size in the matrix data structure in each case a mean distance value ( d ') is determined and this average distance value (d') is assigned to the image point corresponding to its position of the reduced matrix data structure. Method according to one of the preceding claims, characterized in that the viewing area ( 22 ) is set by a) in the height profile (H), in particular in the matrix data structure or in the point cloud, a number of possible viewing ranges (R _1,1 , ...) are predetermined in advance, b) for the respective segment (S ₁ , ..., S _q ) the respective breathing depth (T) on the basis of each of the possible viewing areas ( 22 ), and c) the predetermined possible viewing area (R _1,1 , ...) as the viewing area ( 22 ), for which the determined breathing depth ( 22 ) is the largest. Method according to one of the preceding claims, characterized in that the viewing area ( 22 ) by locating a) in the elevation profile, in particular in the matrix data structure or in the point cloud, by means of object recognition, or by selecting areas corresponding to a human head and torso in advance, and b) the area of the height profile which images the torso , in particular the matrix data structure or the point cloud, or a subarea of this area as a viewing area ( 22 ) is selected, wherein in particular - for detecting the course or the breathing depth of the chest breathing of the head near partial area of this area as a viewing area ( 22 ) is selected and / or - for detecting the course or the breathing depth of the abdominal breathing, the portion of this area remote from the head as a viewing area ( 22 ) is selected. A method according to claim 10, characterized in that for each segment to be examined (S ₁ , ..., S _q ) or for individual height profiles (H) of a segment (S ₁ , ..., S _q ) of the viewing area ( 22 ) by means of object recognition to the height profiles (H) present in the segment (S ₁ , ..., S _q ) starting from the position of the selected viewing region ( 22 ) in the respective preceding segment (S ₁ , ..., S _q ) is adaptively newly determined. Method according to one of the preceding claims, characterized in that the viewing area ( 22 ) is set by a) for all points of the height profile (H), in particular for all entries of the matrix data structure or the point cloud, separately the variance over the respective segment (S ₁ , ..., S _q ) is determined, and b) one area or several areas ( 22a . 22b ) with contiguous points of the height profile (H) whose respective variance is above a lower threshold value or within a predetermined interval, as the viewing area ( 22 ) or, in particular, - by providing related areas ( 22a . 22b ) are selected as points of view whose size exceeds a predetermined number of points - in particular a region is considered to be contiguous, if each point of the height profile can be reached from a point within that region via adjacent pixels, preferably points of the height profile (H) are then considered to be adjacent, - if they are defined by adjacent entries of the matrix data structure, or - if they or their projection on a horizontal plane in space have a threshold below the distance from each other. Method according to claim 12, characterized in that the viewing area ( 22 ) is determined separately in two or more consecutive segments (S ₁ ,..., S _q ) and a height profile (H) is discarded and an error message is issued when the viewing area (FIG. 22 ), in particular its size and / or its center of gravity, relative to the respective preceding segment (S ₁ , ..., S _q ) shifts by a predetermined threshold. Method according to one of the preceding claims, characterized in that a) simultaneously in parallel with the distance of the person from the detector unit ( 20 ) determines the sound emitted by the person and in the form of an audio signal ( 5a ) is provided with a number of audio samples, b) the audio signal (s _a ) is subdivided into a number of audio segments (S _A1 , ..., S _Ar ), in particular with a length of 100 ms to 1 second, and for each of the audio segments (S S _A1 , ..., S _ar ) is examined, whether in this human breath sounds or other sounds are included, and for each audio segment (S _A1 , ..., S _Ar ) each have a classification result is available, c ) simultaneously recorded audio segments (S _A1 , ..., S _ar ) and segments (S ₁ , ..., S _q ) are assigned to each other, and d) audio segments (S _A1 , ..., S _ar ) or segments (S ₁ , ..., S _q ) with low breathing depth or with missing breath sounds is sought and the such an audio segment (S _A1 , ..., S _ar ) or segment (S ₁ , .. ., S _q ) associated segments (S ₁ , ..., S _q ) or audio segments (S _A1 , ..., S _ar ) are also examined for the presence of low breathing depth or lack of breathing sounds, and e) for the Case, that in each other zugeo rdneten segments (S ₁ , ..., S _q ) and audio segments (S ₁ , ..., S _q ) each low breathing depth and lack of breath sounds were detected, the lack of respiration of the person is detected. A method according to claim 14, characterized in that a) a classification signal (A (t)) is created, which contains for each audio segment (S _Ai ) each at least one classification result, the type and strength of the respective noise in the time domain of the respective Audio segment (S _Ai ) indicates, b) that in each case a number of consecutive audio segments (S _Ai ) to other audio segments (S _Ai ), in particular with a duration of 5 to 10 seconds are summarized, preferably the same Time ranges include, like the segments (S ₁ ,..., S _q ), c) an averaging is performed via the classification results contained within a further audio segment (S _Ai ) and the respective average value for the further audio segment (S _Ai ) d) a further classification signal (B (t)) is produced by interpolation of the mean values of the further audio segments (S _Ai ), e) in the further classification signal B (t) and in the respiratory depression signal T (t) is searched for times at which strong changes occur in the two signals (T (t), B (t)), f) the times thus recognized will be considered to be the more relevant, the stronger the change of the respective signal at the respective time , g) a respective relevance value is assigned to these times, and h) those times are detected as starting points (R _S ) or end points (R _E ) of a respiratory arrest for which the magnitude of the relevance value (REL) is above the magnitude of a threshold value, wherein in particular - the threshold is formed by a mean value of the reference measure is formed over a time range, in particular before and / or after the comparison time with the reference dimension and the threshold is set in the range between 120% and 200% of this average. Data carrier on which a program for carrying out a method according to one of the preceding claims is stored. Device for determining the time course of the breathing depth (T) of a person, in particular a sleeping person, comprising - one on a sleeping accommodation, in particular a bed ( 10 ) alignable detector unit ( 20 ), which continuously at successive recording times (t ₁ , ..., t _p ) each have a height profile (H) of the person ( 1 ), wherein in the height profile (H) a number of at least two points in the space is defined, which on the surface of the person ( 1 ) or on the surface of one on or next to the person ( 1 ), and - a processing unit ( 50 ), which a) stores and makes available the respective height profile (H) for each of the recording times (t ₁ ,..., t _p ) in a data structure, b) a number of recording times (t ₁ ,. ., t _p ), in particular within a time range of 3 to 20 seconds, summed up height profiles (H) to form a segment (S ₁ , ..., S _q ), c) a region that covers the abdomen or chest region of the person depending on the respective reference point or reference range ( 21 ) indicates as a viewing area ( 22 ), d) for each height profile (H) within the segment (S ₁ ,..., S _q ) separately the mean value of the distances within the viewing area ( 22 ) points of the height profile (H) to a reference point or reference object, and for the segment (S ₁ , ..., S _q ) a signal (s _i ) is determined to the respective recording time (t ₁ , ..., t _p ) of the height profile (H) is assigned to the mean value determined for this height profile (H), and e) in particular subsequently determines a value characterizing the breathing depth (T) based on this signal (s _i ) or its signal amplitude, characterized in that the processing unit ( 50 f) produces a two-dimensional matrix data structure comprising a number of rows and columns characterizing the height profile (H); g) - predetermining a number of rows arranged in rows and columns in which the distance values (d ₁ ,. .., d _n ) of the height profile (H) are determined, or - the matrix data structure is created by measuring the distance values (d ₁ , ..., d _n ) as the distances along predetermined beams emanating from the detector unit and then calculates the vertical distances from the respective intersection with the surface to the reference plane using the measured distance and the respective measuring angle used, whereby it selects the angles of the various measuring beams so that when the measuring beams hit a plane parallel to the reference plane ( 21 h) where the matrix data structure has a grid of the same size and structure, i) the matrix data structure is created by storing and making available the distance values (d ₁ ,..., d _n ) recorded at the positions in the raster corresponding storage positions in the matrix data structure, and j) creating a reduced matrix data structure, by replacing the matrix data structure after its creation in the dimensions, the spatial resolution of which reduces it, by combining several entries of the matrix data structure into an entry of a reduced matrix data structure, in particular taking only individual ones of the distance measurements (d) for the formation of the reduced matrix data structure, and rejects the remaining distance measurement values, and - preferably determining the parameters (a, b) as integers and determining the reduced matrix data structure H _r (x, y, t) = H (ax, by, t). Device for determining the time course of the breathing depth (T) of a person, in particular a sleeping person, comprising - one on a sleeping accommodation, in particular a bed ( 10 ), alignable detector unit ( 20 ), which continuously at successive recording times (t ₁ , ..., t _p ) each have a height profile (H) of the person ( 1 ), wherein the height profile (H) has a number of at least two distance values (d ₁ , ..., d _n ) for defining a respective point in space, wherein the individual distance values in each case the distance of the intersection of a predetermined relative to the detector unit , in particular from the detector unit emanating beam with the surface of the person ( 1 ) or the surface of one on or next to the person ( 1 ) from a reference point or reference plane ( 21 ) and a processing unit ( 50 ), a) for each of the recording times (t ₁ , ..., t _p ) each creates a data structure containing the respective height profile (H), all of the data structures thus created each having the same size and each with memory positions for the individual Distance values of the height profile (H), b) a number of successive recording times (t ₁ , ..., t _p ), in particular within a time range of 3 to 20 seconds, recorded height profiles (H) to a segment (S ₁ , ..., S _q ), c) a number of memory positions of the data structure in which distance values (d ₁ ,..., D _n ) are stored which are the distance of the abdomen or chest area of the person depending on the respective one Reference point or reference range ( 21 ), as a viewing area ( 22 ), d) for each height profile (H) within a segment (S ₁ ,..., S _q ) the mean value of each within the viewing area (S _{1 ,.} 22 ) distance values (d ₁ , ... d _n ) are determined, and for the segment (S ₁ , ..., S _q ) a signal (s _i ) is determined to which at the respective recording time (t ₁ , ..., t _p) of the height profile (H) is assigned to the mean value (for this height profile H), e) in particular, then based (on this signal, s _i) or its signal amplitude determines a respiratory depth (T) characterizing value, characterized in that Processing unit ( 50 f) produces a two-dimensional matrix data structure comprising a number of rows and columns characterizing the height profile (H); g) - predetermining a number of rows arranged in rows and columns in which the distance values (d ₁ ,. .., d _n ) of the height profile (H) are determined, or - the matrix data structure is created by measuring the distance values (d ₁ , ..., d _n ) as the distances along predetermined beams emanating from the detector unit and then calculates the vertical distances from the respective intersection with the surface to the reference plane using the measured distance and the respective measuring angle used, whereby it selects the angles of the various measuring beams so that when the measuring beams hit a plane parallel to the reference plane ( 21 h) wherein the matrix data structure has a raster of the same size and structure, i) the matrix data structure is created by plotting the distance values (d ₁ , ..., d _n ) recorded at the respective positions and j) create a reduced matrix data structure by replacing the matrix data structure after its creation in the dimensions whose spatial resolution reduces it by adding multiple entries of the matrix data structure to an entry in particular, it uses only individual ones of the distance measurement values (d) for the formation of the reduced matrix data structure and discards the remaining distance measurement values, and preferably determines the parameters (a, b) as integers and the reduced matrix data structure H _r (x, y, t) = H (ax, by, t). Apparatus according to claim 17, characterized in that the detector unit ( 20 ) creates the height profiles (H) in the form of point clouds with a number of at least two points in space, the points on the surface of the person ( 1 ) or on the surface of one on or next to the person ( 1 ) are located. Device according to one of claims 17 to 19, characterized in that the processing unit ( 50 ) for characterizing the breathing depth (T) in a segment (S ₁ , ... S _q ) from the signal (s _i ) at least one maximum (s _max ) and at least one minimum (s _min ) extracted and the at least one difference ( Δ) between the maximum (s _max ) and the minimum (s _min ) as the depth of breath (T) characterizing value for the segment (S ₁ , ..., S _q ) associated time range available and optionally uses. Device according to one of claims 17 to 20, characterized in that the processing unit ( 50 ) the signal (s _i ) undergoes a spectral transformation, in particular Fourier transformation, cosine transformation or wavelet transformation, and searches within a given frequency band, in particular from 0.1 Hz to 1 Hz, for that spectral component with the highest signal derivatives and the signal energy this spectral component is used to characterize the breathing depth (T) in this segment (S ₁ , ..., S _q ). Device according to one of claims 17 to 21, characterized in that the processing unit ( 50 ) subjects the signal (s _i ) associated with a segment (S ₁ , ..., S _q ) to noise filtering after its creation, in particular before the determination of the value characterizing the depth of respiration, in particular by a) signal components having a frequency of more suppressed as 0.5 Hz, and / or b) suppressed DC components. Device according to one of claims 17 to 22, characterized in that the processing unit ( 50 a) for a number of overlapping or non-overlapping segments (S ₁ ,..., S _q ) the depth of respiration (T) is determined separately and / or b) the viewing area ( 22 ) for each segment (S ₁ , ..., S _q ) separately, in particular on the basis of the respective preceding segment (S ₁ ,..., S _q ) 22 ), determined. Device according to one of claims 17 to 23, characterized in that the processing unit ( 50 ) - the height profile (H) characterized by a two-dimensional matrix data structure comprising a number of rows and columns, - a number of arranged in rows and columns grid-like arranged positions is given, at each of which the distance values (d ₁ , ..., d _n ) the height profile (H) determines - the matrix data structure has a raster of the same size and structure, and - creates the matrix data structure by taking the distance values (d ₁ , ..., d _n ) recorded at the respective positions at the positions in the raster stores and makes available corresponding memory positions in the matrix data structure. Device according to claim 24, characterized in that the processing unit ( 50 The matrix data structure after its creation is replaced in its dimensions by a reduced matrix data structure, wherein it determines a mean distance value (d ') for rectangular, in particular the entire matrix data structure covering and non-overlapping image areas of the same size in the matrix data structure and this average distance value (d ') assigns the pixel corresponding to its position to the reduced matrix data structure. Device according to claim 24 or 25, characterized in that the processing unit ( 50 ) the viewing area ( 22 ) by specifying a) in the height profile (H), in particular in the matrix data structure or in the point cloud, a number of possible viewing ranges (R _1,1 , ...), b) for the respective segment ( S ₁ , ..., S _q ) the respective breathing depth (T) on the basis of each of the possible viewing areas ( 22 ) and c) the given possible viewing area (R _1,1 , ...) as the viewing area ( 22 ) and for which the determined depth of breath ( 22 ) is the largest. Device according to one of claims 24 to 26, characterized in that the processing unit ( 50 ) the viewing area ( 22 ) by a) searching for areas in the elevation profile, in particular in the matrix data structure or in the point cloud, by object recognition, or providing means for selecting areas corresponding to a human head and torso, and b) the torso imaging area the height profile, in particular the matrix data structure or the point cloud, or selects a subarea of this area as a viewing area, in particular - for detecting the course or the breathing depth of the chest breathing the partial area of this area close to the head as the viewing area ( 22 ) and / or - for detecting the course or the breathing depth of the abdominal breathing, the portion of this area remote from the head as a viewing area ( 22 ) selects. Device according to claim 27, characterized in that the processing unit ( 50 ) for each segment to be examined (S ₁ , ..., S _q ) or for individual height profiles (H) of a segment (S ₁ , ..., S _q ) of the viewing area ( 22 ) by means of object recognition to the height profiles present in the segment (S ₁ , ..., S _q ) starting from the position of the selected viewing region (H) ( 22 ) in the respective preceding segment (S ₁ , ..., S _q ) adaptively newly determined. Device according to one of claims 24 to 28, characterized in that the processing unit ( 50 ) the viewing area ( 22 ) by determining a) for all points of the height profile (H), in particular for all entries of the matrix data structure, separately the variance of the values over the respective segment (S ₁ , ..., S _q ), and b) a range or several areas ( 22a . 22b ) with contiguous points of the height profile (H) whose respective variance is above a lower threshold value or within a predetermined interval, as the viewing area ( 22 ), in particular - by providing related areas ( 22a . 22b ) is selected as a viewing area whose size exceeds a predetermined number of points - in particular a region is considered as contiguous, if each point of the height profile is accessible from a point within this area on each adjacent pixel, - preferably points of the height profile (H ) are deemed to be adjacent, if they are defined by adjacent entries of the matrix data structure, or if they or their projection on a horizontal plane in space have a threshold below the threshold of one another. Device according to claim 29, characterized in that the processing unit ( 50 ) the viewing area ( 22 ) is determined separately in two or more consecutive segments (S ₁ ,..., S _q ) and discards a height profile (H) and optionally outputs an error message if the viewing area (FIG. 22 ), in particular its size and / or its center of gravity, relative to the respective preceding segment (S ₁ , ..., S _q ) shifts by a predetermined threshold. Device according to one of claims 17 to 30, characterized in that the processing unit ( 50 ) a microphone ( 40 ), which at the same time parallel to the recording of the distance of the person from the person ( 1 ) is provided in the form of an audio signal (s _a ) at its output and this audio signal is supplied to the processing unit, and that the processing unit a) the audio signal (s _a ) in a number of audio segments (S _A1 , ..., S _Ar ), in particular with a length of 100 ms to 1 second, divided and examined for each of the audio segments (S _A1 , ..., S _ar ), whether in this human breath sounds or other sounds are heard, and for each audio segment (S _A1 , ..., S _Ar ) each holds a classification result, c) simultaneously recorded audio segments (S _A1 , ..., S _ar ) and segments (S ₁ , ..., S _q ) assigns each other, and d) audio segments (S _A1 , ..., S _ar ) or segments (S ₁ , ..., S _q ) with low breathing depth or with missing breath sounds and that such an audio segment (S _A1 , ..., S _ar ) or segment (S ₁ , ..., S _q ) associated segments (S ₁ , ..., S _q ) or audio segments (S _A1 , ..., S _ar ) are also examined for the presence of low breathing depth or lack of breathing sounds, and e) in the event that mutually associated Segments (S ₁ , ..., S _q ) and audio segments (S ₁ , ..., S _q ) each low breathing depth and lack of breath sounds were detected, the lack of respiration of the person detects. Apparatus according to claim 31, characterized in that the processing unit ( 50 a) a classification signal (A (t)) is created which contains for each audio segment (S _Ai ) at least one classification result which indicates the type and strength of the respective noise in the time domain of the respective audio segment (S _Ai ), b) in each case a number of successive audio segments (S _Ai ) combined to form further audio segments (S _Ai ), in particular with a duration of 5 to 10 seconds, which preferably comprise the same time ranges as the segments (S ₁ , .. ., S _q ), c) performs an averaging over the classification results contained within a further audio segment (S _Ai ) and assigns the respective mean value to the further audio segment (S _Ai ), d) a further classification signal (B (t) ) by interpolation of the mean values of the further audio segments (S _Ai ), e) in the further classification signal B (t) as well as in the respiratory depression signal T (t) searches for times to which in the two signals (T (t), B ( t)) changes considerably reten, and the timings thus recognized as all the more relevant, the stronger the change of the respective signal at each time point, assigning these times each relevant relevance measure, and f) those times as starting points (R _S ) or endpoints (R _E ) of a respiratory arrest, for which the magnitude of the relevance value (REL) is above the threshold value, wherein it forms, in particular, the threshold value by calculating a mean value of the reference measure over a time range, in particular before and / or after the reference time and the threshold between 120% and 200% of this average.

Images